diff --git a/agent/Dockerfile b/agent/Dockerfile
index cc280daf1..ff66faed0 100644
--- a/agent/Dockerfile
+++ b/agent/Dockerfile
@@ -73,10 +73,13 @@ RUN corepack enable && corepack prepare yarn@stable --activate || corepack enabl
 
 # Install Claude Code CLI (the Python SDK requires this binary)
 # Then update known vulnerable transitive packages where fixed versions exist.
-# Pinned 2.1.191 to match the CLI bundled by claude-agent-sdk 0.2.110 (see
+# Pinned 2.1.301 to match the CLI bundled by claude-agent-sdk 0.2.168 (see
 # agent/pyproject.toml) — the SDK and the on-PATH CLI must agree on the control
-# protocol. This version also has the awsCredentialExport behavior #215 needs:
-# returned creds are cached until 5 min before the JSON's `Expiration`, so an
+# protocol. The Claude 5.5 models the platform defaults to (#953) are rejected
+# by CLI builds older than 2.1.280 ("does not support this model"), so this pin
+# is also what makes the default model invocable at all. It also keeps the
+# awsCredentialExport behavior #215 needs (present since 2.1.176): returned
+# creds are cached until 5 min before the JSON's `Expiration`, so an
 # 8 h task re-assumes the 1 h-capped SessionRole before expiry. Older builds
 # only refreshed hourly on a timer, racing the role-chaining cap.
 # claude-code@2.1.x ships an error-shim at bin/claude.exe that its postinstall
@@ -87,7 +90,7 @@ RUN corepack enable && corepack prepare yarn@stable --activate || corepack enabl
 # after the install so the native binary is placed at build time, and hard-verify
 # it exec's (``claude --version`` fails the build if the shim is still on PATH).
 RUN npm install -g npm@latest && \
-    npm install -g @anthropic-ai/claude-code@2.1.191 && \
+    npm install -g @anthropic-ai/claude-code@2.1.301 && \
     CLAUDE_NPM_ROOT="$(npm root -g)/@anthropic-ai/claude-code" && \
     npm --prefix "${CLAUDE_NPM_ROOT}" update tar minimatch glob cross-spawn picomatch && \
     node "${CLAUDE_NPM_ROOT}/install.cjs" && \
diff --git a/agent/README.md b/agent/README.md
index 46fbb04d1..9386a46e7 100644
--- a/agent/README.md
+++ b/agent/README.md
@@ -119,11 +119,11 @@ The `run.sh` script overrides the container's default CMD to run `python /app/sr
 | `AWS_SECRET_ACCESS_KEY` | Conditional† | | Explicit keys, if you are not using CLI-based resolution |
 | `AWS_SESSION_TOKEN` | No | | For temporary credentials |
 | `AWS_PROFILE` | No | | Profile for `aws configure export-credentials` in `run.sh`, or default profile when using the `~/.aws` mount fallback |
-| `ANTHROPIC_MODEL` | No | `global.anthropic.claude-opus-5` | Bedrock **inference profile** ID for `InvokeModel` (see [inference profiles](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-use.html)). Must be a geo-prefixed profile ID matching the deployment's `bedrockGeoRegion`, not a bare foundation-model ID — see [Model configuration](../docs/guides/DEVELOPER_GUIDE.md#model-configuration) |
+| `ANTHROPIC_MODEL` | No | `global.anthropic.claude-opus-5-5` | Bedrock **inference profile** ID for `InvokeModel` (see [inference profiles](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-use.html)). Must be a geo-prefixed profile ID matching the deployment's `bedrockGeoRegion`, not a bare foundation-model ID — see [Model configuration](../docs/guides/DEVELOPER_GUIDE.md#model-configuration) |
 | `MAX_TURNS` | No | `100` | Max agent turns before stopping |
 | `MAX_BUDGET_USD` | No | | **Local batch only** (shell env when running `entrypoint.py` directly). Range 0.01–100; agent stops when the budget is reached. For deployed AgentCore **server** mode and production tasks, set **`max_budget_usd`** on task creation (REST API, CLI `--max-budget`, or Blueprint default); the orchestrator sends it in the `/invocations` JSON body — server mode does not read `MAX_BUDGET_USD` from the environment. |
 | `DRY_RUN` | No | | Set to `1` to validate config and print the prompt without running the agent |
-| `ANTHROPIC_DEFAULT_HAIKU_MODEL` | No | `global.anthropic.claude-haiku-4-5-20251001-v1:0` | Bedrock **inference profile** ID for the small/fast auxiliary model — the pre-flight safety check and WebFetch summarization (see below). Set by the CDK stack (`cdk/src/stacks/agent.ts` (the runtime environment block)), which derives the prefix from `bedrockGeoRegion` — a geo prefix is required |
+| `ANTHROPIC_DEFAULT_HAIKU_MODEL` | No | `global.anthropic.claude-haiku-5-5` | Bedrock **inference profile** ID for the small/fast auxiliary model — the pre-flight safety check and WebFetch summarization (see below). Set by the CDK stack (`cdk/src/stacks/agent.ts` (the runtime environment block)), which derives the prefix from `bedrockGeoRegion` — a geo prefix is required |
 | `NUDGES_TABLE_NAME` | No | | **Phase 2.** DynamoDB table for mid-task user nudges (`<user_nudge>` XML blocks injected between turns). If unset, the agent runs without nudge support — `nudge_reader.read_pending()` returns `[]` and logs a WARN once. Set automatically by the CDK stack on both AgentCore runtimes. |
 | `JIRA_APP_ACTOR_PROXY_URL` | No | | Resolved per-task from the Jira tenant secret. Forge v2 web-trigger URL used for app-authored Jira comments and transitions. |
 | `JIRA_APP_ACTOR_SHARED_SECRET` | No | | Resolved per-task from the Jira tenant secret. HMAC key for the Forge proxy; redacted from agent diagnostics. |
@@ -133,7 +133,7 @@ The `run.sh` script overrides the container's default CMD to run `python /app/sr
 including non-Jira tasks, so a warm AgentCore process cannot expose one
 tenant's OAuth or Forge credential to the next task.
 
-**Bedrock model access (main model):** Configuring `ANTHROPIC_MODEL` and IAM credentials is not enough. Your AWS account must be able to **invoke** that model in Amazon Bedrock: follow [Request access to models](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html) (Marketplace permissions on first use, Anthropic first-time use where required, valid payment method for Marketplace-backed models). Always use an inference profile ID such as `global.anthropic.claude-opus-5`: a bare foundation-model ID cannot be invoked with on-demand throughput and Bedrock rejects it with `ValidationException`. IAM must also grant the model — see [Model configuration](../docs/guides/DEVELOPER_GUIDE.md#model-configuration) for the full layering. If the CLI stops with a message that the model is not available on your Bedrock deployment, fix model access in the console or switch `ANTHROPIC_MODEL` to an entitled profile, then retry.
+**Bedrock model access (main model):** Configuring `ANTHROPIC_MODEL` and IAM credentials is not enough. Your AWS account must be able to **invoke** that model in Amazon Bedrock: follow [Request access to models](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html) (Marketplace permissions on first use, Anthropic first-time use where required, valid payment method for Marketplace-backed models). Always use an inference profile ID such as `global.anthropic.claude-opus-5-5`: a bare foundation-model ID cannot be invoked with on-demand throughput and Bedrock rejects it with `ValidationException`. IAM must also grant the model — see [Model configuration](../docs/guides/DEVELOPER_GUIDE.md#model-configuration) for the full layering. If the CLI stops with a message that the model is not available on your Bedrock deployment, fix model access in the console or switch `ANTHROPIC_MODEL` to an entitled profile, then retry.
 
 **Pre-flight check model**: Claude Code runs a quick safety verification using a small Haiku model before executing each tool command. On Bedrock, the default Haiku model ID may not be enabled in your account, causing the check to time out with *"Pre-flight check is taking longer than expected"* warnings. The agent sets `ANTHROPIC_DEFAULT_HAIKU_MODEL` to a known-available Bedrock Haiku model ID to avoid this. If you see pre-flight timeout warnings, verify that this model is enabled in your Bedrock model access settings.
 
@@ -145,7 +145,7 @@ tenant's OAuth or Forge credential to the next task.
 # Dry run — validate config, fetch issue, print assembled prompt, then exit
 DRY_RUN=1 ./agent/run.sh "owner/repo" 42
 
-# Run with a specific model (overrides the global.anthropic.claude-opus-5 default).
+# Run with a specific model (overrides the global.anthropic.claude-opus-5-5 default).
 # Must be a geo-prefixed inference profile that IAM grants — see Model configuration.
 ANTHROPIC_MODEL="us.anthropic.claude-sonnet-4-6" ./agent/run.sh "owner/repo" 42
 
diff --git a/agent/pyproject.toml b/agent/pyproject.toml
index 40894cb5b..4f3741b03 100644
--- a/agent/pyproject.toml
+++ b/agent/pyproject.toml
@@ -16,12 +16,12 @@ dependencies = [
     # would degrade gracefully — but for now we keep the dep to
     # preserve the clean code path.
     "bedrock-agentcore==1.18.1", #https://pypi.org/project/bedrock-agentcore/
-    "claude-agent-sdk==0.2.110", #https://github.com/anthropics/claude-agent-sdk-python/releases/tag/v0.2.110 (bundles claude CLI 2.1.191; kept in lockstep with the npm CLI pin in the Dockerfile, #215)
+    "claude-agent-sdk==0.2.168", #https://github.com/anthropics/claude-agent-sdk-python/releases/tag/v0.2.168 (bundles claude CLI 2.1.301; kept in lockstep with the npm CLI pin in the Dockerfile, #215. Claude 5.5 models need CLI >= 2.1.280, #953)
     "requests==2.34.2", #https://pypi.org/project/requests/
     "fastapi==0.139.0", #https://pypi.org/project/fastapi/
     "uvicorn==0.50.0", #https://pypi.org/project/uvicorn/
     "aws-opentelemetry-distro==0.18.0", #https://pypi.org/project/aws-opentelemetry-distro/
-    "mcp==1.28.1", #https://pypi.org/project/mcp/
+    "mcp==1.30.0", #https://pypi.org/project/mcp/ (1.30.0 clears GHSA-84m7-p3x7-pcfv)
     # CEDAR ENGINE PARITY — DO NOT BUMP IN ISOLATION.
     # cedarpy (Python, agent runtime) and @cedar-policy/cedar-wasm (TypeScript,
     # CDK Lambdas) are two language bindings over the same Cedar Rust core.
diff --git a/agent/scripts/diagnostics/test_sdk_smoke.py b/agent/scripts/diagnostics/test_sdk_smoke.py
index b3e905a0f..0a9328f10 100644
--- a/agent/scripts/diagnostics/test_sdk_smoke.py
+++ b/agent/scripts/diagnostics/test_sdk_smoke.py
@@ -39,7 +39,7 @@ async def smoke_test():
     if not model:
         print(
             "ERROR: ANTHROPIC_MODEL not set. Pass the geo-prefixed inference-profile "
-            "id the deployment uses, e.g. ANTHROPIC_MODEL=global.anthropic.claude-opus-5",
+            "id the deployment uses, e.g. ANTHROPIC_MODEL=global.anthropic.claude-opus-5-5",
             file=sys.stderr,
         )
         sys.exit(1)
diff --git a/agent/src/config.py b/agent/src/config.py
index 3e1c635e1..8cbc79cae 100644
--- a/agent/src/config.py
+++ b/agent/src/config.py
@@ -694,7 +694,7 @@ def build_config(
     resolved_github_token = github_token or resolve_github_token()
     resolved_aws_region = aws_region or os.environ.get("AWS_REGION", "")
     resolved_anthropic_model = anthropic_model or os.environ.get(
-        "ANTHROPIC_MODEL", "global.anthropic.claude-opus-5"
+        "ANTHROPIC_MODEL", "global.anthropic.claude-opus-5-5"
     )
     # Small/fast auxiliary model (WebFetch summarization etc.). Falls back to the
     # deployed ANTHROPIC_DEFAULT_HAIKU_MODEL env, then the platform default. Must
@@ -705,7 +705,7 @@ def build_config(
     # env set at all reaches this literal, so the two must be moved together or a
     # local run silently uses a different geography than a deployed one.
     resolved_haiku_model = haiku_model or os.environ.get(
-        "ANTHROPIC_DEFAULT_HAIKU_MODEL", "global.anthropic.claude-haiku-4-5-20251001-v1:0"
+        "ANTHROPIC_DEFAULT_HAIKU_MODEL", "global.anthropic.claude-haiku-5-5"
     )
 
     # Resolve the workflow id (the create-task boundary already pinned it; local
diff --git a/agent/src/models.py b/agent/src/models.py
index 8c170f97c..393fcb749 100644
--- a/agent/src/models.py
+++ b/agent/src/models.py
@@ -154,7 +154,7 @@ class TaskConfig(BaseModel):
     task_description: str = ""
     github_token: str = ""
     aws_region: str
-    anthropic_model: str = "global.anthropic.claude-opus-5"
+    anthropic_model: str = "global.anthropic.claude-opus-5-5"
     # The "small/fast" model Claude Code uses for auxiliary work (e.g. WebFetch
     # page summarization). Must be a cross-region INFERENCE-PROFILE id (geo
     # prefix), not a bare foundation-model id — Claude 4.x cannot be invoked
@@ -164,7 +164,7 @@ class TaskConfig(BaseModel):
     # vars from its resolved bedrockGeoRegion. It applies to direct construction
     # (tests, local runs), so it must name the same geography as the deployment or
     # those paths silently exercise a different one.
-    haiku_model: str = "global.anthropic.claude-haiku-4-5-20251001-v1:0"
+    haiku_model: str = "global.anthropic.claude-haiku-5-5"
     dry_run: bool = False
     max_turns: int = 10
     max_budget_usd: float | None = None
diff --git a/agent/tests/test_config.py b/agent/tests/test_config.py
index 12b42233c..662eb9fcc 100644
--- a/agent/tests/test_config.py
+++ b/agent/tests/test_config.py
@@ -63,7 +63,7 @@ def test_haiku_model_defaults_to_inference_profile(self, monkeypatch):
             github_token="ghp_test123",
             aws_region="us-east-1",
         )
-        assert config.haiku_model == "global.anthropic.claude-haiku-4-5-20251001-v1:0"
+        assert config.haiku_model == "global.anthropic.claude-haiku-5-5"
 
     def test_haiku_model_resolves_from_env(self, monkeypatch):
         # The deployed ANTHROPIC_DEFAULT_HAIKU_MODEL (set by agent.ts) flows
diff --git a/agent/uv.lock b/agent/uv.lock
index 82c9f9414..0e9849726 100644
--- a/agent/uv.lock
+++ b/agent/uv.lock
@@ -171,10 +171,10 @@ requires-dist = [
     { name = "bedrock-agentcore", specifier = "==1.18.1" },
     { name = "boto3", specifier = "==1.43.78" },
     { name = "cedarpy", specifier = "==4.8.4" },
-    { name = "claude-agent-sdk", specifier = "==0.2.110" },
+    { name = "claude-agent-sdk", specifier = "==0.2.168" },
     { name = "fastapi", specifier = "==0.139.0" },
     { name = "jsonschema", specifier = "==4.26.0" },
-    { name = "mcp", specifier = "==1.28.1" },
+    { name = "mcp", specifier = "==1.30.0" },
     { name = "pyyaml", specifier = "==6.0.3" },
     { name = "requests", specifier = "==2.34.2" },
     { name = "uvicorn", specifier = "==0.50.0" },
@@ -399,20 +399,21 @@ wheels = [
 
 [[package]]
 name = "claude-agent-sdk"
-version = "0.2.110"
+version = "0.2.168"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "anyio" },
+    { name = "jsonschema" },
     { name = "mcp" },
     { name = "sniffio" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/bb/98/8fdab35ed9e1a36bc7afab4d390cc5002094a4950996c079da9aa4541cc4/claude_agent_sdk-0.2.110.tar.gz", hash = "sha256:538b548bac07a22f65686abab063a902ac76ba35989d0f073c942f96248e9fa3", size = 255632, upload-time = "2026-06-24T22:11:52.342Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/9a/ee/31e159daf956f0c5503eac67d02743da692c0999fef046cdde3cc52702b9/claude_agent_sdk-0.2.168.tar.gz", hash = "sha256:80388054a09c137d76fc29caa6e46bc90d2ae2ef09fac5ee1c38fc36d2762fda", size = 371591, upload-time = "2026-10-15T20:20:43.683Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/aa/93/29d4fdaa13e69034faf8d3503df915b07c820e2c08e3d6a7515149cde5bb/claude_agent_sdk-0.2.110-py3-none-macosx_11_0_arm64.whl", hash = "sha256:fed0e0f4804d9f9cff80ab7d1b44142ebd1046cdd29ca74caef4c92c35fff8d8", size = 64924533, upload-time = "2026-06-24T22:11:55.612Z" },
-    { url = "https://files.pythonhosted.org/packages/aa/03/b40bb673cd93cdc3928262c1be75fde34a7bed4bf2c2c20e04218e2005ea/claude_agent_sdk-0.2.110-py3-none-macosx_11_0_x86_64.whl", hash = "sha256:62b23869d46cef6f6ff1d00ceaa5e846e2f1d297478421c835efb8fe99369d4f", size = 69704449, upload-time = "2026-06-24T22:11:59.149Z" },
-    { url = "https://files.pythonhosted.org/packages/f9/18/ab67cb5ce641333385bed55ed8e9665c00f7d30d1f6ab12f8463ddb7695f/claude_agent_sdk-0.2.110-py3-none-manylinux_2_17_aarch64.whl", hash = "sha256:324e49553c6303d6b267217dc2912652b97af2bc96503efd12095ae915b46b83", size = 74879555, upload-time = "2026-06-24T22:12:03.25Z" },
-    { url = "https://files.pythonhosted.org/packages/91/88/3627d7d14310cfec66977551263e219365244a906fc7ca1209fb0c3a6cec/claude_agent_sdk-0.2.110-py3-none-manylinux_2_17_x86_64.whl", hash = "sha256:56371dd7a2c66c0bd497dc0b3cab4193a228b196f600676393d69c0ecee37cfb", size = 75924237, upload-time = "2026-06-24T22:12:07.183Z" },
-    { url = "https://files.pythonhosted.org/packages/49/79/c9066c5c387d42c19a4b675ec1ff5219f8920cfda8ff8b527119fd69b774/claude_agent_sdk-0.2.110-py3-none-win_amd64.whl", hash = "sha256:4235d4de6d685a189c12612095ab192b759280ede1f3aed0c3e784d52c3555f9", size = 75448209, upload-time = "2026-06-24T22:12:11.283Z" },
+    { url = "https://files.pythonhosted.org/packages/15/31/9a205fc0b7ae1b66d5964774bb534ab378047c9bf87578f443bd1fb2205e/claude_agent_sdk-0.2.168-py3-none-macosx_11_0_arm64.whl", hash = "sha256:2ef1e7fc442c5abdc1e698027bf1b5faa8ea2130134a14364e2ae3c53bf31fdd", size = 104355641, upload-time = "2026-10-15T20:20:48.105Z" },
+    { url = "https://files.pythonhosted.org/packages/9f/ba/12bbe92d5e9145c7aaa94aa349a0859b77a9b58ccc40ae9e69cf7f078e5b/claude_agent_sdk-0.2.168-py3-none-macosx_11_0_x86_64.whl", hash = "sha256:653fc7bbe56630e6ddd9601a7cfe50586667e74102da3fb90571456c63655a62", size = 109438860, upload-time = "2026-10-15T20:20:52.274Z" },
+    { url = "https://files.pythonhosted.org/packages/ad/5a/26e587846f14f67f3353f96811b86931263a9c7e264d17a691d9577e4da5/claude_agent_sdk-0.2.168-py3-none-manylinux_2_17_aarch64.whl", hash = "sha256:77ae66b10edfd07b6ea3dec2af106ad47a9d616f8eb52d4f9b385e1b00bb2a27", size = 114030032, upload-time = "2026-10-15T20:20:56.813Z" },
+    { url = "https://files.pythonhosted.org/packages/00/03/09d347dcddbbab4becd41fef3a811967eddf24135b2d62dac9e321c81537/claude_agent_sdk-0.2.168-py3-none-manylinux_2_17_x86_64.whl", hash = "sha256:be197be92549909ec62a2cee04c27f4b0a38c80f723735538a3efc352507f81d", size = 114443626, upload-time = "2026-10-15T20:21:01.794Z" },
+    { url = "https://files.pythonhosted.org/packages/48/93/520a18ab853b688afcc31d504344e541a7dc68bc2b515ea8a5fef1c868a1/claude_agent_sdk-0.2.168-py3-none-win_amd64.whl", hash = "sha256:ed1814b0dd1c113bdaf3e51e6a67aa5b96b8f5c5839a43aeb6ae1e6ed23a5b35", size = 117292475, upload-time = "2026-10-15T20:21:07.053Z" },
 ]
 
 [[package]]
@@ -701,7 +702,7 @@ wheels = [
 
 [[package]]
 name = "mcp"
-version = "1.28.1"
+version = "1.30.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "anyio" },
@@ -719,9 +720,9 @@ dependencies = [
     { name = "typing-inspection" },
     { name = "uvicorn", marker = "sys_platform != 'emscripten'" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/6e/77/9450b8f251a13affb6281997d0523c4615f8a8b35d0b21ff30db3a5aac9d/mcp-1.28.1.tar.gz", hash = "sha256:d51e36a5f5644faea4f85ea649bfffa6bc6c26770d42798ad6a3de3d2ba69683", size = 638501, upload-time = "2026-06-26T12:57:29.093Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/ba/93/0142dc84a666daf8ad51a34268f34c12fd6fda4f3810c4be2504eecc8212/mcp-1.30.0.tar.gz", hash = "sha256:445414625fce5c295faa505bb11bacece661ab6f4028d57c935db57820b7a3e4", size = 680511, upload-time = "2026-09-07T14:34:15.845Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/e2/5e/d118fce19f87a2e7d8101c35c8ae0ec289098a4df0ff244cec23e415aca0/mcp-1.28.1-py3-none-any.whl", hash = "sha256:2726bca5e7193f61c5dde8b12500a6de2d9acf6d1a1c0be9e8c2e706437991df", size = 222620, upload-time = "2026-06-26T12:57:27.218Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/f4/e58bc33317c92a0203664daaf00bf6f41166cc0149e5d6870a03f7cd004a/mcp-1.30.0-py3-none-any.whl", hash = "sha256:666edb5009503e1047c9d60346a756f94b261f05cc2625f23d41c728ffc484d0", size = 234581, upload-time = "2026-09-07T14:34:14.266Z" },
 ]
 
 [[package]]
diff --git a/cdk/src/handlers/shared/bedrock-model-constants.ts b/cdk/src/handlers/shared/bedrock-model-constants.ts
index 36b6ca990..f2a323fe9 100644
--- a/cdk/src/handlers/shared/bedrock-model-constants.ts
+++ b/cdk/src/handlers/shared/bedrock-model-constants.ts
@@ -60,7 +60,7 @@
  * that list means the granted model and the delivered model id cannot drift — a
  * mismatch would AccessDenied every Haiku call at run time while synth stayed green.
  */
-export const DEFAULT_HAIKU_MODEL_ID = 'anthropic.claude-haiku-4-5-20251001-v1:0';
+export const DEFAULT_HAIKU_MODEL_ID = 'anthropic.claude-haiku-5-5';
 
 /**
  * Single source of truth for the Bedrock **foundation-model IDs** the agent runtime
@@ -81,16 +81,23 @@ export const DEFAULT_HAIKU_MODEL_ID = 'anthropic.claude-haiku-4-5-20251001-v1:0'
  * and was removed; `bgagent platform doctor` now checks the whole set for it.
  */
 export const DEFAULT_BEDROCK_MODEL_IDS: readonly string[] = [
+  // Previous-generation models stay granted after a default bump (#744, #953):
+  // blueprints may pin any of them per-repo, so removing one would fail those
+  // repos at turn 0. Retiring a model is a separate, announced change — never a
+  // side effect of adding a newer one.
   'anthropic.claude-sonnet-4-6',
-  // Claude Opus 4.8. Blueprints may pin it per-repo, so removing it would fail those
-  // repos at turn 0.
   'anthropic.claude-opus-4-8',
-  // Claude Opus 5 — the platform default. Bare ID by contract: Bedrock refuses the
-  // bare ID for on-demand invocation ("ValidationException: … isn't supported. Retry
-  // your request with the ID or ARN of an inference profile"), and the grant sites
-  // derive the geo-prefixed inference-profile ARN — the invocable one — from this
-  // entry plus `bedrockGeoRegion`.
   'anthropic.claude-opus-5',
+  'anthropic.claude-haiku-4-5-20251001-v1:0',
+  // Claude Opus 5.5 (platform default) and Haiku 5.5 (auxiliary default), #953.
+  // Bare IDs by contract: Bedrock refuses the bare ID for on-demand invocation
+  // ("ValidationException: … isn't supported. Retry your request with the ID or
+  // ARN of an inference profile"), and the grant sites derive the geo-prefixed
+  // inference-profile ARN — the invocable one — from each entry plus
+  // `bedrockGeoRegion`. Sonnet 5.5 is deliberately NOT here: it has no `jp.`
+  // profile (verified 2026-10-16), so granting it would recreate the
+  // granted-but-uninvocable state described above for one geography.
+  'anthropic.claude-opus-5-5',
   DEFAULT_HAIKU_MODEL_ID,
 ];
 
@@ -106,7 +113,7 @@ export const DEFAULT_BEDROCK_MODEL_IDS: readonly string[] = [
  * The auxiliary default is an ALIAS of {@link DEFAULT_HAIKU_MODEL_ID}, not a second
  * copy of the string: they must name the same model, and two literals could drift.
  */
-export const PLATFORM_DEFAULT_MODEL_ID = 'anthropic.claude-opus-5';
+export const PLATFORM_DEFAULT_MODEL_ID = 'anthropic.claude-opus-5-5';
 export const PLATFORM_DEFAULT_AUX_MODEL_ID = DEFAULT_HAIKU_MODEL_ID;
 
 /**
diff --git a/cdk/test/constructs/ecs-agent-cluster.test.ts b/cdk/test/constructs/ecs-agent-cluster.test.ts
index 3c5a95688..98eafdf89 100644
--- a/cdk/test/constructs/ecs-agent-cluster.test.ts
+++ b/cdk/test/constructs/ecs-agent-cluster.test.ts
@@ -435,8 +435,16 @@ describe('EcsAgentCluster construct', () => {
     // on-demand invocable, so the `us.` profile is the one actually called.
     expect(serialized).toContain('foundation-model/anthropic.claude-opus-5');
     expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-5');
-    // REGRESSION (#744): Opus 4.8 stays granted alongside Opus 5 — blueprints may
-    // pin it per-repo, so dropping it would fail those repos at turn 0.
+    // Claude 5.5 family (#953) — the platform default (Opus 5.5) and auxiliary
+    // default (Haiku 5.5). Both ARNs each: the bare id is not on-demand invocable,
+    // so the `us.`-prefixed inference profile is the one actually called.
+    expect(serialized).toContain('foundation-model/anthropic.claude-opus-5-5');
+    expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-5-5');
+    expect(serialized).toContain('foundation-model/anthropic.claude-haiku-5-5');
+    expect(serialized).toContain('inference-profile/us.anthropic.claude-haiku-5-5');
+    // REGRESSION (#744, #953): previous defaults stay granted after a bump.
+    // Blueprints may pin any of them per-repo; dropping one would fail those repos
+    // at turn 0.
     expect(serialized).toContain('foundation-model/anthropic.claude-opus-4-8');
     expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-4-8');
   });
@@ -456,15 +464,26 @@ describe('EcsAgentCluster construct', () => {
   test('default-context Bedrock grants are the exact expected set, nothing wider', () => {
     const EXPECTED_BEDROCK_RESOURCE_NAMES = [
       'foundation-model/anthropic.claude-haiku-4-5-20251001-v1:0',
+      'foundation-model/anthropic.claude-haiku-5-5',
       'foundation-model/anthropic.claude-opus-4-8',
       'foundation-model/anthropic.claude-opus-5',
+      'foundation-model/anthropic.claude-opus-5-5',
       'foundation-model/anthropic.claude-sonnet-4-6',
       'inference-profile/us.anthropic.claude-haiku-4-5-20251001-v1:0',
+      'inference-profile/us.anthropic.claude-haiku-5-5',
       'inference-profile/us.anthropic.claude-opus-4-8',
       'inference-profile/us.anthropic.claude-opus-5',
+      'inference-profile/us.anthropic.claude-opus-5-5',
       'inference-profile/us.anthropic.claude-sonnet-4-6',
     ];
-    const serialized = JSON.stringify(baseTemplate.findResources('AWS::IAM::Policy'));
+    // Inline AND managed-policy overflow, matching the AgentCore half: CDK splits a
+    // role's default policy into managed-policy chunks past the IAM size limit, and
+    // a grant that spilled over must still count here rather than vanish from view.
+    const policies = {
+      ...baseTemplate.findResources('AWS::IAM::Policy'),
+      ...baseTemplate.findResources('AWS::IAM::ManagedPolicy'),
+    };
+    const serialized = JSON.stringify(Object.values(policies).map(policy => policy.Properties.PolicyDocument));
     const found = [...new Set(
       serialized.match(/(?:foundation-model|inference-profile)\/[^"]+/g) ?? [],
     )].sort();
@@ -497,9 +516,9 @@ describe('EcsAgentCluster construct', () => {
       expect(containers.length).toBeGreaterThan(0);
       for (const c of containers) {
         const env = Object.fromEntries((c.Environment ?? []).map((e) => [e.Name, e.Value]));
-        expect(env.ANTHROPIC_MODEL).toBe(`${geo}.anthropic.claude-opus-5`);
+        expect(env.ANTHROPIC_MODEL).toBe(`${geo}.anthropic.claude-opus-5-5`);
         expect(env.ANTHROPIC_DEFAULT_HAIKU_MODEL)
-          .toBe(`${geo}.anthropic.claude-haiku-4-5-20251001-v1:0`);
+          .toBe(`${geo}.anthropic.claude-haiku-5-5`);
       }
     }
   });
diff --git a/cdk/test/stacks/agent.test.ts b/cdk/test/stacks/agent.test.ts
index dedd325e0..6b55bc154 100644
--- a/cdk/test/stacks/agent.test.ts
+++ b/cdk/test/stacks/agent.test.ts
@@ -266,7 +266,7 @@ describe('AgentStack', () => {
       const envVars = (rt as { Properties?: { EnvironmentVariables?: Record<string, unknown> } })
         .Properties?.EnvironmentVariables ?? {};
       expect(envVars.ANTHROPIC_DEFAULT_HAIKU_MODEL)
-        .toBe(`${DEFAULT_BEDROCK_GEO_REGION}.anthropic.claude-haiku-4-5-20251001-v1:0`);
+        .toBe(`${DEFAULT_BEDROCK_GEO_REGION}.anthropic.claude-haiku-5-5`);
     }
   });
 
@@ -387,9 +387,16 @@ describe('AgentStack', () => {
     // must be present or the agent gets AccessDenied at turn 0.
     expect(serialized).toContain('foundation-model/anthropic.claude-opus-5');
     expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-5');
-    // REGRESSION (#744): Opus 4.8 stays granted alongside Opus 5. Blueprints may
-    // pin 4.8 per-repo; dropping it would fail those repos at turn 0. Retiring
-    // 4.8 is a separate, announced change — not a side effect of adding 5.
+    // Claude 5.5 family (#953) — the platform default (Opus 5.5) and auxiliary
+    // default (Haiku 5.5). Both ARNs each: the bare id is not on-demand invocable,
+    // so the `us.`-prefixed inference profile is the one actually called.
+    expect(serialized).toContain('foundation-model/anthropic.claude-opus-5-5');
+    expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-5-5');
+    expect(serialized).toContain('foundation-model/anthropic.claude-haiku-5-5');
+    expect(serialized).toContain('inference-profile/us.anthropic.claude-haiku-5-5');
+    // REGRESSION (#744, #953): previous defaults stay granted after a bump.
+    // Blueprints may pin any of them per-repo; dropping one would fail those repos
+    // at turn 0. Retiring a model is a separate, announced change.
     expect(serialized).toContain('foundation-model/anthropic.claude-opus-4-8');
     expect(serialized).toContain('inference-profile/us.anthropic.claude-opus-4-8');
   });
@@ -407,8 +414,8 @@ describe('AgentStack', () => {
       context: {
         bedrockModels: [
           'anthropic.claude-opus-4-8',
-          'anthropic.claude-opus-5',
-          'anthropic.claude-haiku-4-5-20251001-v1:0',
+          'anthropic.claude-opus-5-5',
+          'anthropic.claude-haiku-5-5',
         ],
       },
     });
@@ -496,12 +503,16 @@ describe('AgentStack', () => {
     // added, dropped, or re-prefixed without failing here.
     const EXPECTED_BEDROCK_RESOURCE_NAMES = [
       'foundation-model/anthropic.claude-haiku-4-5-20251001-v1:0',
+      'foundation-model/anthropic.claude-haiku-5-5',
       'foundation-model/anthropic.claude-opus-4-8',
       'foundation-model/anthropic.claude-opus-5',
+      'foundation-model/anthropic.claude-opus-5-5',
       'foundation-model/anthropic.claude-sonnet-4-6',
       'inference-profile/us.anthropic.claude-haiku-4-5-20251001-v1:0',
+      'inference-profile/us.anthropic.claude-haiku-5-5',
       'inference-profile/us.anthropic.claude-opus-4-8',
       'inference-profile/us.anthropic.claude-opus-5',
+      'inference-profile/us.anthropic.claude-opus-5-5',
       'inference-profile/us.anthropic.claude-sonnet-4-6',
     ];
 
@@ -528,7 +539,7 @@ describe('AgentStack', () => {
       Properties?: { EnvironmentVariables?: Record<string, unknown> };
     }).Properties?.EnvironmentVariables ?? {};
     expect(envVars.ANTHROPIC_DEFAULT_HAIKU_MODEL)
-      .toBe('us.anthropic.claude-haiku-4-5-20251001-v1:0');
+      .toBe('us.anthropic.claude-haiku-5-5');
     expect(DEFAULT_BEDROCK_GEO_REGION).toBe('us');
   });
 
@@ -566,9 +577,9 @@ describe('AgentStack', () => {
       );
       expect(envs.length).toBeGreaterThan(0);
       for (const env of envs) {
-        expect(env.ANTHROPIC_MODEL).toBe(`${geo}.anthropic.claude-opus-5`);
+        expect(env.ANTHROPIC_MODEL).toBe(`${geo}.anthropic.claude-opus-5-5`);
         expect(env.ANTHROPIC_DEFAULT_HAIKU_MODEL)
-          .toBe(`${geo}.anthropic.claude-haiku-4-5-20251001-v1:0`);
+          .toBe(`${geo}.anthropic.claude-haiku-5-5`);
       }
     });
 
@@ -611,7 +622,7 @@ describe('AgentStack', () => {
         const envVars = (rt as { Properties?: { EnvironmentVariables?: Record<string, unknown> } })
           .Properties?.EnvironmentVariables ?? {};
         expect(envVars.ANTHROPIC_DEFAULT_HAIKU_MODEL)
-          .toBe(`${geo}.anthropic.claude-haiku-4-5-20251001-v1:0`);
+          .toBe(`${geo}.anthropic.claude-haiku-5-5`);
       }
     });
   });
@@ -1094,7 +1105,7 @@ describe('AgentStack', () => {
  * independent of `bedrockModels`, so a narrowing override made the two disagree with
  * nothing to catch it. Verified before fixing: `-c
  * bedrockModels='["anthropic.claude-sonnet-4-6"]'` synthesized clean while granting only
- * Sonnet and injecting `global.anthropic.claude-opus-5`.
+ * Sonnet and injecting `global.anthropic.claude-opus-5-5`.
  */
 describe('AgentStack platform-default / grant-list coherence', () => {
   const synth = (bedrockModels?: string[]) => () => {
@@ -1109,8 +1120,8 @@ describe('AgentStack platform-default / grant-list coherence', () => {
     const t = synth(['anthropic.claude-sonnet-4-6']);
     expect(t).toThrow(/omits the platform default model/);
     // Names every missing default, not just the first one found.
-    expect(t).toThrow(/anthropic\.claude-opus-5/);
-    expect(t).toThrow(/claude-haiku-4-5/);
+    expect(t).toThrow(/anthropic\.claude-opus-5-5/);
+    expect(t).toThrow(/claude-haiku-5-5/);
     // And says WHY, because the override looks self-consistent on its own.
     expect(t).toThrow(/told to invoke them regardless/);
   });
@@ -1118,14 +1129,14 @@ describe('AgentStack platform-default / grant-list coherence', () => {
   test('fails when only the AUXILIARY default is omitted', () => {
     // The easy one to miss: adding a model while dropping Haiku breaks only the
     // auxiliary path (WebFetch summarisation), which no task-level failure names.
-    expect(synth(['anthropic.claude-opus-5', 'anthropic.claude-sonnet-4-6']))
-      .toThrow(/claude-haiku-4-5/);
+    expect(synth(['anthropic.claude-opus-5-5', 'anthropic.claude-sonnet-4-6']))
+      .toThrow(/claude-haiku-5-5/);
   });
 
   test('accepts an override that keeps both defaults', () => {
     expect(synth([
-      'anthropic.claude-opus-5',
-      'anthropic.claude-haiku-4-5-20251001-v1:0',
+      'anthropic.claude-opus-5-5',
+      'anthropic.claude-haiku-5-5',
       'anthropic.claude-sonnet-4-6',
     ])).not.toThrow();
   });
diff --git a/cdk/test/stacks/compute-selection.test.ts b/cdk/test/stacks/compute-selection.test.ts
index 1237110fb..a7182bb0f 100644
--- a/cdk/test/stacks/compute-selection.test.ts
+++ b/cdk/test/stacks/compute-selection.test.ts
@@ -121,7 +121,14 @@ describe.each(['agentcore', 'ecs', 'lambda-microvm'])('exclusive %s deployment',
     expect(trust.includes('LambdaMicrovmComputeExecutionRole')).toBe(backend === 'lambda-microvm');
     expect(trust).toContain('sts:TagSession');
     const prefix = backend === 'agentcore' ? 'RuntimeExecutionRole' : backend === 'ecs' ? 'EcsAgentClusterTaskRole' : 'LambdaMicrovmComputeExecutionRole';
-    const policies = JSON.stringify(Object.entries(template.findResources('AWS::IAM::Policy')).filter(([id]) => id.startsWith(prefix)));
+    // Audit the role's inline policy AND its managed-policy overflow. CDK splits a
+    // role's default policy into `AWS::IAM::ManagedPolicy` chunks once the inline
+    // document exceeds the IAM size limit; the Bedrock grant list (#953) pushed the
+    // AgentCore runtime role past it, so a statement can legitimately live in either.
+    const policies = JSON.stringify([
+      ...Object.entries(template.findResources('AWS::IAM::Policy')),
+      ...Object.entries(template.findResources('AWS::IAM::ManagedPolicy')),
+    ].filter(([id]) => id.startsWith(prefix)));
     expect(policies).toContain('bedrock-agentcore:InvokeGateway');
     expect(policies).toContain('bedrock-agentcore:GetResourceOauth2Token');
   });
diff --git a/cli/src/model-id.ts b/cli/src/model-id.ts
index 72925cd2a..652e98f91 100644
--- a/cli/src/model-id.ts
+++ b/cli/src/model-id.ts
@@ -106,7 +106,7 @@ export function assertModelIdUsable(args: {
   if (!bare) {
     throw new CliError(
       `--model '${modelId}' is a geography prefix with no model after it. Use `
-      + `'${geo}.<model-id>', e.g. '${geo}.anthropic.claude-opus-5'.`,
+      + `'${geo}.<model-id>', e.g. '${geo}.anthropic.claude-opus-5-5'.`,
     );
   }
   // A second geo prefix on the bare half means the operator prefixed twice; the
diff --git a/cli/src/repo-display.ts b/cli/src/repo-display.ts
index 768656ab8..629953991 100644
--- a/cli/src/repo-display.ts
+++ b/cli/src/repo-display.ts
@@ -52,7 +52,7 @@ export const PLATFORM_REPO_DEFAULTS = {
    * if it names a different model than the runtime invokes, doctor can report a
    * healthy stack while every task fails at turn 0 with AccessDenied.
    */
-  model_id: 'global.anthropic.claude-opus-5',
+  model_id: 'global.anthropic.claude-opus-5-5',
   max_turns: 200,
   poll_interval_ms: 30_000,
   approval_gate_cap: 50,
diff --git a/cli/test/platform-doctor.test.ts b/cli/test/platform-doctor.test.ts
index 3956cf98b..2bb18b10d 100644
--- a/cli/test/platform-doctor.test.ts
+++ b/cli/test/platform-doctor.test.ts
@@ -500,14 +500,14 @@ describe('doctor Bedrock inference-profile check', () => {
     const send = jest.fn().mockResolvedValue({});
     const check = await profileCheck('global', send);
     expect(check.status).toBe('pass');
-    expect(check.label).toContain('global.anthropic.claude-opus-5');
+    expect(check.label).toContain('global.anthropic.claude-opus-5-5');
 
     // The queried identifier is the geo-prefixed profile, not the bare model id.
     const queried = send.mock.calls
       .map(([c]) => (c as { _type?: string; input?: { inferenceProfileIdentifier?: string } }))
       .filter((c) => c._type === 'GetInferenceProfile')
       .map((c) => c.input?.inferenceProfileIdentifier);
-    expect(queried).toContain('global.anthropic.claude-opus-5');
+    expect(queried).toContain('global.anthropic.claude-opus-5-5');
   });
 
   it('uses the geography the stack reports, not a hardcoded one', async () => {
@@ -515,7 +515,7 @@ describe('doctor Bedrock inference-profile check', () => {
     // `us` and must be checked against `us.`, not against the current default.
     const send = jest.fn().mockResolvedValue({});
     const check = await profileCheck('us', send);
-    expect(check.label).toContain('us.anthropic.claude-opus-5');
+    expect(check.label).toContain('us.anthropic.claude-opus-5-5');
     expect(check.label).not.toContain('global.');
   });
 
diff --git a/docs/abca-plugin/skills/onboard-repo/SKILL.md b/docs/abca-plugin/skills/onboard-repo/SKILL.md
index 3333025cb..26162f08d 100644
--- a/docs/abca-plugin/skills/onboard-repo/SKILL.md
+++ b/docs/abca-plugin/skills/onboard-repo/SKILL.md
@@ -40,9 +40,9 @@ Use AskUserQuestion to collect (only the repository is required — the rest fal
 
 - **Repository** — GitHub `owner/repo`. Must match exactly what's passed to `bgagent submit --repo` later.
 - **Compute type** — `agentcore` (default) or `ecs`.
-- **Model** — default is the platform model (Opus 5). The geo prefix in the examples
+- **Model** — default is the platform model (Opus 5.5). The geo prefix in the examples
   below (`global.`) must match the deployment's `bedrockGeoRegion`; `bgagent repo
-  onboard` rejects a mismatch at the CLI rather than letting the task fail at turn 0. If overriding, it must be a model **already granted to the runtime** (see "Model not yet wired into the runtime"), specified as a cross-Region **inference-profile ID** (e.g. `global.anthropic.claude-opus-5`), not a raw `anthropic.*` foundation-model ID.
+  onboard` rejects a mismatch at the CLI rather than letting the task fail at turn 0. If overriding, it must be a model **already granted to the runtime** (see "Model not yet wired into the runtime"), specified as a cross-Region **inference-profile ID** (e.g. `global.anthropic.claude-opus-5-5`), not a raw `anthropic.*` foundation-model ID.
 - **Max turns** — default 100 (range 1–500).
 - **Per-repo GitHub token** — only if this repo needs a different token than the platform default (provide its Secrets Manager ARN).
 
@@ -59,7 +59,7 @@ no `cdk deploy`.**
 ```bash
 bgagent repo onboard <owner/repo>
 # common overrides:
-#   --model <inference-profile-id>     e.g. global.anthropic.claude-opus-5 (must be runtime-granted)
+#   --model <inference-profile-id>     e.g. global.anthropic.claude-opus-5-5 (must be runtime-granted)
 #   --compute-type <agentcore|ecs>
 #   --max-turns <n>                    per-repo default turn limit
 #   --token-secret-arn <arn>           per-repo GitHub token (else platform default)
@@ -79,7 +79,7 @@ That's it — the repo is onboarded. Submit a task with the `submit-task` skill.
 **Pick a model that is already wired into the runtime.** With no `--model`, the repo
 uses the platform default — read it from the stack's `BedrockGeoRegion` +
 `BedrockModelIds` outputs rather than a literal here. If you pass `--model`, use a cross-Region
-**inference profile ID** (e.g. `global.anthropic.claude-opus-5`), not a raw
+**inference profile ID** (e.g. `global.anthropic.claude-opus-5-5`), not a raw
 `anthropic.*` foundation-model ID. Only models the stack has granted the runtime can
 be invoked — see "Model not yet wired into the runtime" before choosing a model the
 deployment doesn't already support.
@@ -100,7 +100,7 @@ editing the stack and redeploying.
      repoTable: repoTable.table,
      // Optional overrides:
      // computeType: 'agentcore',
-     // modelId: 'global.anthropic.claude-opus-5',
+     // modelId: 'global.anthropic.claude-opus-5-5',
      // maxTurns: 100,
      // maxBudgetUsd: 50,
      // githubTokenSecretArn: 'arn:aws:secretsmanager:...',
@@ -118,8 +118,8 @@ editing the stack and redeploying.
 
 A repo can only use a model the **runtime IAM role has `grantInvoke` for**. The granted
 set is `DEFAULT_BEDROCK_MODEL_IDS` in
-`cdk/src/handlers/shared/bedrock-model-constants.ts` — Sonnet 4.6,
-Opus 4.8, Opus 5, and Haiku 4.5 — and a deployed stack publishes it as the
+`cdk/src/handlers/shared/bedrock-model-constants.ts` — Opus 5.5 and Haiku 5.5 (the
+defaults) plus Sonnet 4.6, Opus 4.8, Opus 5, and Haiku 4.5 — and a deployed stack publishes it as the
 `BedrockModelIds` output, so read that rather than trusting this list to stay current:
 
 ```bash
@@ -138,20 +138,20 @@ ids for on-demand invocation, and the IAM grant is scoped to one geography's pro
 Granting a **new** model is a deploy-time change, not a construct edit: the list is overridable via CDK context, either on the command line or in `cdk.json`. Both forms behave identically — the resolver JSON-parses the string `-c` delivers.
 
 ```bash
-cdk deploy -c bedrockModels='["anthropic.claude-opus-5","anthropic.claude-haiku-4-5-20251001-v1:0","anthropic.claude-sonnet-4-6"]'
+cdk deploy -c bedrockModels='["anthropic.claude-opus-5-5","anthropic.claude-haiku-5-5","anthropic.claude-sonnet-4-6"]'
 ```
 
-The override **REPLACES** the default list rather than adding to it, so it must include the two models the stack injects as `ANTHROPIC_MODEL` and `ANTHROPIC_DEFAULT_HAIKU_MODEL` — Opus 5 and Haiku 4.5. Omitting either fails at synth, naming the missing model, because every substrate is told to invoke them regardless of this list. (`--context-file` does not work for this: it is accepted and silently ignored.)
+The override **REPLACES** the default list rather than adding to it, so it must include the two models the stack injects as `ANTHROPIC_MODEL` and `ANTHROPIC_DEFAULT_HAIKU_MODEL` — Opus 5.5 and Haiku 5.5. Omitting either fails at synth, naming the missing model, because every substrate is told to invoke them regardless of this list. (`--context-file` does not work for this: it is accepted and silently ignored.)
 
 For a persistent setting, put the same array in the `context` block of `cdk.json`:
 
 ```jsonc
 // cdk.json — this REPLACES the default list, so include the platform defaults
-// (Opus 5 + Haiku 4.5) or the stack's own defaults are ungranted
+// (Opus 5.5 + Haiku 5.5) or the stack's own defaults are ungranted
 "context": {
   "bedrockModels": [
-    "anthropic.claude-opus-5",
-    "anthropic.claude-haiku-4-5-20251001-v1:0",
+    "anthropic.claude-opus-5-5",
+    "anthropic.claude-haiku-5-5",
     "anthropic.claude-sonnet-4-6"
   ]
 }
@@ -186,7 +186,7 @@ platform default, and treat "add model X" as a separate, later change.
 |---------|---------|---------|
 | `compute_type` | Execution strategy | `agentcore` |
 | `runtime_arn` | AgentCore runtime override | Platform default |
-| `model_id` | AI model for tasks (inference profile ID) | Platform default (Opus 5, as `<BedrockGeoRegion>.anthropic.claude-opus-5`) |
+| `model_id` | AI model for tasks (inference profile ID) | Platform default (Opus 5.5, as `<BedrockGeoRegion>.anthropic.claude-opus-5-5`) |
 | `max_turns` | Turn limit per task | 100 |
 | `max_budget_usd` | Cost ceiling per task | Unlimited |
 | `system_prompt_overrides` | Custom system instructions | None |
@@ -199,6 +199,6 @@ Task-level parameters override per-repo defaults; if neither specifies a value,
 
 - **`REPO_NOT_ONBOARDED` / 422** — the repo isn't registered. Run `bgagent repo onboard <owner/repo>` (Path A). Confirm the `owner/repo` matches exactly what you pass to `bgagent submit --repo`.
 - **Preflight failure after onboarding** — the GitHub PAT lacks access to the new repo. Ensure the token has Contents (read/write) + Pull requests (read/write) on it, or onboard with a repo-specific `--token-secret-arn`.
-- **400 "Invocation with on-demand throughput isn't supported"** — `model_id` is a raw foundation-model ID; use the inference-profile ID (e.g. `global.anthropic.claude-opus-5`).
+- **400 "Invocation with on-demand throughput isn't supported"** — `model_id` is a raw foundation-model ID; use the inference-profile ID (e.g. `global.anthropic.claude-opus-5-5`).
 - **403 "not authorized to perform bedrock:InvokeModelWithResponseStream"** — the repo's model isn't wired into the runtime. See "Model not yet wired into the runtime."
 - **Model not available / "not available on your Bedrock deployment"** — account-level Bedrock access isn't enabled for that model/Region (separate from IAM); complete [model access](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html), then use an enabled inference-profile ID.
diff --git a/docs/abca-plugin/skills/troubleshoot/SKILL.md b/docs/abca-plugin/skills/troubleshoot/SKILL.md
index 7a6d3805e..ef96c3d17 100644
--- a/docs/abca-plugin/skills/troubleshoot/SKILL.md
+++ b/docs/abca-plugin/skills/troubleshoot/SKILL.md
@@ -111,13 +111,13 @@ node cli/lib/bin/bgagent.js events <TASK_ID> --output json
 - Common: repo build/test commands not documented in CLAUDE.md
 
 **403 "not authorized to perform bedrock:InvokeModelWithResponseStream":**
-- The repo's `model_id` is a model the runtime IAM role wasn't granted. The runtime only has `grantInvoke` for the models in the stack's configured set — read it from the `BedrockModelIds` stack output rather than a list here (Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5 by default).
-- **Quick fix:** point the repo at an already-granted model — `bgagent repo onboard <owner/repo> --model global.anthropic.claude-opus-5` (no redeploy).
-- **To add a new model to the runtime:** grant it in the stack and redeploy. The model set is the shared list in `cdk/src/constructs/bedrock-models.ts` — add the model via the `bedrockModels` CDK context (`cdk.json`) so both the AgentCore and ECS backends grant it (#433). Adding a model also requires **account-level Bedrock access** for it (separate from IAM — see the next row).
+- The repo's `model_id` is a model the runtime IAM role wasn't granted. The runtime only has `grantInvoke` for the models in the stack's configured set — read it from the `BedrockModelIds` stack output rather than a list here (Opus 5.5, Haiku 5.5, Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5 by default).
+- **Quick fix:** point the repo at an already-granted model — `bgagent repo onboard <owner/repo> --model global.anthropic.claude-opus-5-5` (no redeploy).
+- **To add a new model to the runtime:** grant it in the stack and redeploy. The model set is `DEFAULT_BEDROCK_MODEL_IDS` in `cdk/src/handlers/shared/bedrock-model-constants.ts` — add the model via the `bedrockModels` CDK context (`cdk.json`) so both the AgentCore and ECS backends grant it (#433). Adding a model also requires **account-level Bedrock access** for it (separate from IAM — see the next row).
 
 **Model not enabled / "not available on your Bedrock deployment" (often immediate failure, few turns, zero or near-zero tokens):**
 - **IAM is necessary but not sufficient.** The AgentCore role may already have `bedrock:InvokeModel*`, but the **account** must also satisfy [Amazon Bedrock model access](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html): Marketplace subscription flow on first serverless use (with `aws-marketplace:Subscribe` / `ViewSubscriptions` where needed), Anthropic **first-time use** details (`PutUseCaseForModelAccess` or the console model catalog), and a valid payment method for Marketplace-backed models.
-- **Use an inference profile ID** in the Blueprint / DynamoDB `model_id` when Bedrock requires it for on-demand invocation (for example `global.anthropic.claude-opus-5` for global Opus 5). See [Use an inference profile in model invocation](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-use.html). Raw `anthropic.*` IDs often hit "on-demand not supported" or wrong routing — see the **400** row below.
+- **Use an inference profile ID** in the Blueprint / DynamoDB `model_id` when Bedrock requires it for on-demand invocation (for example `global.anthropic.claude-opus-5-5` for global Opus 5.5). See [Use an inference profile in model invocation](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-use.html). Raw `anthropic.*` IDs often hit "on-demand not supported" or wrong routing — see the **400** row below.
 - **Cross-Region profiles** route across Regions in a geography; ensure IAM and any SCPs allow Bedrock in **all destination Regions** for that profile. See [Supported Regions and models for inference profiles](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-support.html).
 - **Task status:** When the Claude CLI reports a terminal error via `ResultMessage.is_error`, the agent marks the task **FAILED** (not COMPLETED) and persists `error_message` in DynamoDB.
 
diff --git a/docs/design/ARCHITECTURE.md b/docs/design/ARCHITECTURE.md
index f73387750..b046b0c3d 100644
--- a/docs/design/ARCHITECTURE.md
+++ b/docs/design/ARCHITECTURE.md
@@ -62,8 +62,8 @@ Different tasks and repos may benefit from different models. The `model_id` fiel
 |---|---|---|---|
 | `coding/new-task-v1` | Claude Sonnet 4 | `anthropic.claude-sonnet-4-6` | Good balance of quality and cost |
 | `coding/pr-iteration-v1` | Claude Sonnet 4 | `anthropic.claude-sonnet-4-6` | Needs to understand review feedback and make code changes |
-| `coding/pr-review-v1` | Claude Haiku | `anthropic.claude-haiku-4-5-20251001-v1:0` | Fast and cheap - review is read-only analysis |
-| Complex/critical repos | Claude Opus 4.8 | `anthropic.claude-opus-4-8` | Highest quality, opt-in per repo |
+| `coding/pr-review-v1` | Claude Haiku 5.5 | `anthropic.claude-haiku-5-5` | Fast and cheap - review is read-only analysis |
+| Complex/critical repos | Claude Opus 5.5 | `anthropic.claude-opus-5-5` | Highest quality; also the platform default when a repo pins nothing |
 
 ## Cost model
 
diff --git a/docs/design/BEDROCK_COST_ATTRIBUTION.md b/docs/design/BEDROCK_COST_ATTRIBUTION.md
index 1358461e8..a161f1c97 100644
--- a/docs/design/BEDROCK_COST_ATTRIBUTION.md
+++ b/docs/design/BEDROCK_COST_ATTRIBUTION.md
@@ -86,7 +86,7 @@ Set via the process env (not project settings) so the untrusted repo can't alter
 
 ## Version alignment
 
-The agent runs Claude Code two ways that must agree on the control protocol: the `claude-agent-sdk` Python wheel **bundles** a CLI, and the Dockerfile also installs the CLI via npm. Both are pinned in lockstep — `claude-agent-sdk==0.2.110` (bundles CLI 2.1.191) and npm `@anthropic-ai/claude-code@2.1.191`. 2.1.191 also satisfies the ≥2.1.176 `awsCredentialExport`-with-`Expiration` requirement.
+The agent runs Claude Code two ways that must agree on the control protocol: the `claude-agent-sdk` Python wheel **bundles** a CLI, and the Dockerfile also installs the CLI via npm. Both are pinned in lockstep — `claude-agent-sdk==0.2.168` (bundles CLI 2.1.301) and npm `@anthropic-ai/claude-code@2.1.301`. 2.1.301 also satisfies the ≥2.1.176 `awsCredentialExport`-with-`Expiration` requirement and the ≥2.1.280 floor the Claude 5.5 default models need.
 
 ## Track 3 — operator documentation
 
diff --git a/docs/design/INTERACTIVE_AGENTS.md b/docs/design/INTERACTIVE_AGENTS.md
index c17955d97..cb09255c1 100644
--- a/docs/design/INTERACTIVE_AGENTS.md
+++ b/docs/design/INTERACTIVE_AGENTS.md
@@ -766,12 +766,12 @@ Opt-in per task: 4 KB previews + full trajectory to S3 with TTL.
 
 ## Appendix A — Claude Agent SDK reference
 
-Pinned version: `claude-agent-sdk==0.2.110` (Python; see `agent/pyproject.toml`).
+Pinned version: `claude-agent-sdk==0.2.168` (Python; see `agent/pyproject.toml`).
 
 > **Caution — re-verify against 0.2.x.** The hook-surface details in this
 > appendix (the `HookEvent` enum members, `PostToolUseFailure`, the Stop-hook
 > return-value contract, etc.) were originally written against
-> `claude-agent-sdk==0.1.53`. The pin has since advanced to `0.2.110`, and the
+> `claude-agent-sdk==0.1.53`. The pin has since advanced to `0.2.168`, and the
 > SDK's hook API may have changed across that range. Before relying on any
 > specific enum member or hook signature below, verify it against the installed
 > 0.2.x SDK and the actual usage in `agent/src/hooks.py`.
diff --git a/docs/design/REPO_ONBOARDING.md b/docs/design/REPO_ONBOARDING.md
index 3a2ca7ab2..7bbc3855a 100644
--- a/docs/design/REPO_ONBOARDING.md
+++ b/docs/design/REPO_ONBOARDING.md
@@ -120,7 +120,7 @@ From lowest to highest priority:
 |---|---|---|
 | `compute_type` | First deployed backend (`agentcore` with legacy context) | Ordered `DEPLOYED_COMPUTE_TYPE` list on the orchestrator; `ComputeTypes`, `ComputeSubstrate` and `ComputeDeploymentMode` outputs for CLI discovery |
 | `runtime_arn` | Stack-level env var | CDK stack props |
-| `model_id` | `global.anthropic.claude-opus-5` | injected by the stack as `ANTHROPIC_MODEL` from `bedrockGeoRegion`; `agent/src/config.py` holds the no-env fallback — see [Model configuration](../guides/DEVELOPER_GUIDE.md#model-configuration) |
+| `model_id` | `global.anthropic.claude-opus-5-5` | injected by the stack as `ANTHROPIC_MODEL` from `bedrockGeoRegion`; `agent/src/config.py` holds the no-env fallback — see [Model configuration](../guides/DEVELOPER_GUIDE.md#model-configuration) |
 | `max_turns` | 100 | Platform constant |
 | `max_budget_usd` | None (unlimited) | No platform default by design — a global ceiling would kill long tasks mid-change. Set a per-repo default with Blueprint `agent.maxBudgetUsd` (`0.01`–`100`, validated at synth) or per task with `--max-budget` / `max_budget_usd`. See [Per-repo overrides](../guides/USER_GUIDE.md#per-repo-overrides) for the complete list of surfaces a budget can come from |
 | `memory_token_budget` | 2000 | Platform constant |
diff --git a/docs/guides/DEPLOYMENT_GUIDE.md b/docs/guides/DEPLOYMENT_GUIDE.md
index e2f4997dc..4bd7c31fb 100644
--- a/docs/guides/DEPLOYMENT_GUIDE.md
+++ b/docs/guides/DEPLOYMENT_GUIDE.md
@@ -213,7 +213,7 @@ At public US East (N. Virginia) first-tier list rates verified in August 2026, t
 
 | Service | Used By | Scales to Zero |
 |---------|---------|---------------|
-| Bedrock (Claude Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5) | Agent reasoning, cross-region inference profiles | Yes |
+| Bedrock (Claude Opus 5.5, Haiku 5.5, plus Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5 for pinned repos) | Agent reasoning, cross-region inference profiles | Yes |
 | Bedrock Guardrails | Prompt injection detection on task input | Yes |
 | Bedrock AgentCore Memory | Semantic + episodic extraction strategies | Yes |
 | AWS Agent Registry (preview, default-on) | Versioned agent asset catalog and governance | N/A (managed preview service) |
diff --git a/docs/guides/DEVELOPER_GUIDE.md b/docs/guides/DEVELOPER_GUIDE.md
index 10cbce3cd..d977aadfa 100644
--- a/docs/guides/DEVELOPER_GUIDE.md
+++ b/docs/guides/DEVELOPER_GUIDE.md
@@ -76,7 +76,7 @@ definitions.push({
   repo: 'acme/my-service',
   compute: { runtimeArn: '...' },                    // override the default runtime ARN
   agent: {
-    modelId: 'global.anthropic.claude-opus-5',       // foundation model override
+    modelId: 'global.anthropic.claude-opus-5-5',     // foundation model override
     maxTurns: 150,                                    // default turn limit for this repo
     systemPromptOverrides: 'Extra instructions...',   // appended to the platform prompt
   },
@@ -160,7 +160,7 @@ See the [Cedar policy guide](./CEDAR_POLICY_GUIDE.md) for the full authoring ref
 per-task payload model_id            (layer 5)
   > blueprint agent.modelId          (layer 4, arrives as stack env ANTHROPIC_MODEL)
   > stack env ANTHROPIC_MODEL        (layer 3-adjacent / local shell)
-  > agent/src/config.py fallback     (layer 2 — global.anthropic.claude-opus-5)
+  > agent/src/config.py fallback     (layer 2 — global.anthropic.claude-opus-5-5)
 ```
 
 Every one of those is gated by the **IAM invoke allowlist** (layer 1), which is itself gated by **account-level Bedrock model access**. Both gates are silent until invocation: a model that resolves fine through precedence still fails at turn 0 with `AccessDenied` if it is not in the grant list, and fails again if your account has not completed [Bedrock model access](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html) for it.
@@ -178,7 +178,7 @@ throughput isn't supported. Retry your request with the ID or ARN of an inferenc
 profile that contains this model.
 ```
 
-So: `bedrockModels` context → `anthropic.claude-opus-5`. Everywhere else → `global.anthropic.claude-opus-5` (or whatever geo you have configured — see below).
+So: `bedrockModels` context → `anthropic.claude-opus-5-5`. Everywhere else → `global.anthropic.claude-opus-5-5` (or whatever geo you have configured — see below).
 
 ### Choosing the inference-profile geography
 
@@ -209,6 +209,7 @@ Changing the geography requires a **redeploy**, because the IAM grants are scope
 4. Update the **prefixed** ID in `agent/src/config.py` and `agent/src/models.py` too. These are the fallback a run with no env reaches (a local batch run), and the doc-drift test compares against them.
 5. **Verify the SDK price table recognizes the model.** The `max_budget_usd` guardrail is computed from a price table bundled into the Claude Agent SDK at build time, so an unrecognized model silently degrades budget enforcement. Run `agent/scripts/diagnostics/test_sdk_smoke.py` with `ANTHROPIC_MODEL` set to the new ID, divide the reported cost by the input-token count, and confirm the implied rate matches [published Bedrock pricing](https://aws.amazon.com/bedrock/pricing/). A `$0.00` or wildly-off result means the table does not know the model and budgets cannot be trusted.
 6. The doc-drift test (`cdk/test/contracts/model-default-docs-parity.test.ts`) fails until the documented defaults here and in `agent/README.md` match `config.py`. That failure is the reminder, not a nuisance — update both.
+7. **Tell operators who override `bedrockModels`.** The override REPLACES the default list, and synth fails ("omits the platform default model") if it lacks either new default. A deployment carrying `-c bedrockModels=[...]` or a `cdk.json` `context` entry must add the new bare IDs before pulling the bump — call this out in the release notes.
 
 ### Cost and model selection
 
@@ -378,7 +379,7 @@ The `--local-events` flag connects the agent container to DynamoDB Local on the
 
 | Variable | Default | Description |
 |---|---|---|
-| `ANTHROPIC_MODEL` | `global.anthropic.claude-opus-5` | Bedrock inference-profile ID for the main coding model |
+| `ANTHROPIC_MODEL` | `global.anthropic.claude-opus-5-5` | Bedrock inference-profile ID for the main coding model |
 | `MAX_TURNS` | `100` | Max agent turns before stopping |
 | `MAX_BUDGET_USD` | | Cost ceiling for local batch runs only (production uses the API field) |
 | `DRY_RUN` | | Set to `1` to validate and print prompt without running the agent |
diff --git a/docs/guides/USER_GUIDE.md b/docs/guides/USER_GUIDE.md
index 7d55b39e5..ea6e2b8c6 100644
--- a/docs/guides/USER_GUIDE.md
+++ b/docs/guides/USER_GUIDE.md
@@ -229,7 +229,7 @@ Blueprints can configure per-repository settings that override platform defaults
 |---|---|---|
 | `compute_type` | Compute strategy (`agentcore`, `ecs`, or `lambda-microvm`) | `agentcore` |
 | `runtime_arn` | AgentCore runtime ARN override | Platform default |
-| `model_id` | Bedrock inference-profile ID (geo-prefixed, matching the deployment's `bedrockGeoRegion`) | `global.anthropic.claude-opus-5` |
+| `model_id` | Bedrock inference-profile ID (geo-prefixed, matching the deployment's `bedrockGeoRegion`) | `global.anthropic.claude-opus-5-5` |
 | `max_turns` | Default turn limit for tasks | 100 |
 | `max_budget_usd` | Default cost budget in USD per task, `0.01`–`100` (Blueprint `agent.maxBudgetUsd`) | None (unlimited) |
 | `system_prompt_overrides` | Additional system prompt instructions | None |
diff --git a/docs/src/content/docs/architecture/Architecture.md b/docs/src/content/docs/architecture/Architecture.md
index 88a176092..7928b8e51 100644
--- a/docs/src/content/docs/architecture/Architecture.md
+++ b/docs/src/content/docs/architecture/Architecture.md
@@ -66,8 +66,8 @@ Different tasks and repos may benefit from different models. The `model_id` fiel
 |---|---|---|---|
 | `coding/new-task-v1` | Claude Sonnet 4 | `anthropic.claude-sonnet-4-6` | Good balance of quality and cost |
 | `coding/pr-iteration-v1` | Claude Sonnet 4 | `anthropic.claude-sonnet-4-6` | Needs to understand review feedback and make code changes |
-| `coding/pr-review-v1` | Claude Haiku | `anthropic.claude-haiku-4-5-20251001-v1:0` | Fast and cheap - review is read-only analysis |
-| Complex/critical repos | Claude Opus 4.8 | `anthropic.claude-opus-4-8` | Highest quality, opt-in per repo |
+| `coding/pr-review-v1` | Claude Haiku 5.5 | `anthropic.claude-haiku-5-5` | Fast and cheap - review is read-only analysis |
+| Complex/critical repos | Claude Opus 5.5 | `anthropic.claude-opus-5-5` | Highest quality; also the platform default when a repo pins nothing |
 
 ## Cost model
 
diff --git a/docs/src/content/docs/architecture/Bedrock-cost-attribution.md b/docs/src/content/docs/architecture/Bedrock-cost-attribution.md
index dbd3a7121..294ff3017 100644
--- a/docs/src/content/docs/architecture/Bedrock-cost-attribution.md
+++ b/docs/src/content/docs/architecture/Bedrock-cost-attribution.md
@@ -90,7 +90,7 @@ Set via the process env (not project settings) so the untrusted repo can't alter
 
 ## Version alignment
 
-The agent runs Claude Code two ways that must agree on the control protocol: the `claude-agent-sdk` Python wheel **bundles** a CLI, and the Dockerfile also installs the CLI via npm. Both are pinned in lockstep — `claude-agent-sdk==0.2.110` (bundles CLI 2.1.191) and npm `@anthropic-ai/claude-code@2.1.191`. 2.1.191 also satisfies the ≥2.1.176 `awsCredentialExport`-with-`Expiration` requirement.
+The agent runs Claude Code two ways that must agree on the control protocol: the `claude-agent-sdk` Python wheel **bundles** a CLI, and the Dockerfile also installs the CLI via npm. Both are pinned in lockstep — `claude-agent-sdk==0.2.168` (bundles CLI 2.1.301) and npm `@anthropic-ai/claude-code@2.1.301`. 2.1.301 also satisfies the ≥2.1.176 `awsCredentialExport`-with-`Expiration` requirement and the ≥2.1.280 floor the Claude 5.5 default models need.
 
 ## Track 3 — operator documentation
 
diff --git a/docs/src/content/docs/architecture/Interactive-agents.md b/docs/src/content/docs/architecture/Interactive-agents.md
index f968bc830..3978af2cc 100644
--- a/docs/src/content/docs/architecture/Interactive-agents.md
+++ b/docs/src/content/docs/architecture/Interactive-agents.md
@@ -770,12 +770,12 @@ Opt-in per task: 4 KB previews + full trajectory to S3 with TTL.
 
 ## Appendix A — Claude Agent SDK reference
 
-Pinned version: `claude-agent-sdk==0.2.110` (Python; see `agent/pyproject.toml`).
+Pinned version: `claude-agent-sdk==0.2.168` (Python; see `agent/pyproject.toml`).
 
 > **Caution — re-verify against 0.2.x.** The hook-surface details in this
 > appendix (the `HookEvent` enum members, `PostToolUseFailure`, the Stop-hook
 > return-value contract, etc.) were originally written against
-> `claude-agent-sdk==0.1.53`. The pin has since advanced to `0.2.110`, and the
+> `claude-agent-sdk==0.1.53`. The pin has since advanced to `0.2.168`, and the
 > SDK's hook API may have changed across that range. Before relying on any
 > specific enum member or hook signature below, verify it against the installed
 > 0.2.x SDK and the actual usage in `agent/src/hooks.py`.
diff --git a/docs/src/content/docs/architecture/Repo-onboarding.md b/docs/src/content/docs/architecture/Repo-onboarding.md
index 80ef5373c..d9c3c6233 100644
--- a/docs/src/content/docs/architecture/Repo-onboarding.md
+++ b/docs/src/content/docs/architecture/Repo-onboarding.md
@@ -124,7 +124,7 @@ From lowest to highest priority:
 |---|---|---|
 | `compute_type` | First deployed backend (`agentcore` with legacy context) | Ordered `DEPLOYED_COMPUTE_TYPE` list on the orchestrator; `ComputeTypes`, `ComputeSubstrate` and `ComputeDeploymentMode` outputs for CLI discovery |
 | `runtime_arn` | Stack-level env var | CDK stack props |
-| `model_id` | `global.anthropic.claude-opus-5` | injected by the stack as `ANTHROPIC_MODEL` from `bedrockGeoRegion`; `agent/src/config.py` holds the no-env fallback — see [Model configuration](/sample-autonomous-cloud-coding-agents/developer-guide/model-configuration) |
+| `model_id` | `global.anthropic.claude-opus-5-5` | injected by the stack as `ANTHROPIC_MODEL` from `bedrockGeoRegion`; `agent/src/config.py` holds the no-env fallback — see [Model configuration](/sample-autonomous-cloud-coding-agents/developer-guide/model-configuration) |
 | `max_turns` | 100 | Platform constant |
 | `max_budget_usd` | None (unlimited) | No platform default by design — a global ceiling would kill long tasks mid-change. Set a per-repo default with Blueprint `agent.maxBudgetUsd` (`0.01`–`100`, validated at synth) or per task with `--max-budget` / `max_budget_usd`. See [Per-repo overrides](/sample-autonomous-cloud-coding-agents/customizing/per-repo-overrides) for the complete list of surfaces a budget can come from |
 | `memory_token_budget` | 2000 | Platform constant |
diff --git a/docs/src/content/docs/customizing/Per-repo-overrides.md b/docs/src/content/docs/customizing/Per-repo-overrides.md
index a2efae5af..cb18a168a 100644
--- a/docs/src/content/docs/customizing/Per-repo-overrides.md
+++ b/docs/src/content/docs/customizing/Per-repo-overrides.md
@@ -8,7 +8,7 @@ Blueprints can configure per-repository settings that override platform defaults
 |---|---|---|
 | `compute_type` | Compute strategy (`agentcore`, `ecs`, or `lambda-microvm`) | `agentcore` |
 | `runtime_arn` | AgentCore runtime ARN override | Platform default |
-| `model_id` | Bedrock inference-profile ID (geo-prefixed, matching the deployment's `bedrockGeoRegion`) | `global.anthropic.claude-opus-5` |
+| `model_id` | Bedrock inference-profile ID (geo-prefixed, matching the deployment's `bedrockGeoRegion`) | `global.anthropic.claude-opus-5-5` |
 | `max_turns` | Default turn limit for tasks | 100 |
 | `max_budget_usd` | Default cost budget in USD per task, `0.01`–`100` (Blueprint `agent.maxBudgetUsd`) | None (unlimited) |
 | `system_prompt_overrides` | Additional system prompt instructions | None |
diff --git a/docs/src/content/docs/developer-guide/Installation.md b/docs/src/content/docs/developer-guide/Installation.md
index 5a01813c6..e0c6579e3 100644
--- a/docs/src/content/docs/developer-guide/Installation.md
+++ b/docs/src/content/docs/developer-guide/Installation.md
@@ -133,7 +133,7 @@ The `--local-events` flag connects the agent container to DynamoDB Local on the
 
 | Variable | Default | Description |
 |---|---|---|
-| `ANTHROPIC_MODEL` | `global.anthropic.claude-opus-5` | Bedrock inference-profile ID for the main coding model |
+| `ANTHROPIC_MODEL` | `global.anthropic.claude-opus-5-5` | Bedrock inference-profile ID for the main coding model |
 | `MAX_TURNS` | `100` | Max agent turns before stopping |
 | `MAX_BUDGET_USD` | | Cost ceiling for local batch runs only (production uses the API field) |
 | `DRY_RUN` | | Set to `1` to validate and print prompt without running the agent |
diff --git a/docs/src/content/docs/developer-guide/Model-configuration.md b/docs/src/content/docs/developer-guide/Model-configuration.md
index 92454140a..48058fd4a 100644
--- a/docs/src/content/docs/developer-guide/Model-configuration.md
+++ b/docs/src/content/docs/developer-guide/Model-configuration.md
@@ -28,7 +28,7 @@ title: Model configuration
 per-task payload model_id            (layer 5)
   > blueprint agent.modelId          (layer 4, arrives as stack env ANTHROPIC_MODEL)
   > stack env ANTHROPIC_MODEL        (layer 3-adjacent / local shell)
-  > agent/src/config.py fallback     (layer 2 — global.anthropic.claude-opus-5)
+  > agent/src/config.py fallback     (layer 2 — global.anthropic.claude-opus-5-5)
 ```
 
 Every one of those is gated by the **IAM invoke allowlist** (layer 1), which is itself gated by **account-level Bedrock model access**. Both gates are silent until invocation: a model that resolves fine through precedence still fails at turn 0 with `AccessDenied` if it is not in the grant list, and fails again if your account has not completed [Bedrock model access](https://docs.aws.amazon.com/bedrock/latest/userguide/model-access.html) for it.
@@ -46,7 +46,7 @@ throughput isn't supported. Retry your request with the ID or ARN of an inferenc
 profile that contains this model.
 ```
 
-So: `bedrockModels` context → `anthropic.claude-opus-5`. Everywhere else → `global.anthropic.claude-opus-5` (or whatever geo you have configured — see below).
+So: `bedrockModels` context → `anthropic.claude-opus-5-5`. Everywhere else → `global.anthropic.claude-opus-5-5` (or whatever geo you have configured — see below).
 
 ### Choosing the inference-profile geography
 
@@ -77,6 +77,7 @@ Changing the geography requires a **redeploy**, because the IAM grants are scope
 4. Update the **prefixed** ID in `agent/src/config.py` and `agent/src/models.py` too. These are the fallback a run with no env reaches (a local batch run), and the doc-drift test compares against them.
 5. **Verify the SDK price table recognizes the model.** The `max_budget_usd` guardrail is computed from a price table bundled into the Claude Agent SDK at build time, so an unrecognized model silently degrades budget enforcement. Run `agent/scripts/diagnostics/test_sdk_smoke.py` with `ANTHROPIC_MODEL` set to the new ID, divide the reported cost by the input-token count, and confirm the implied rate matches [published Bedrock pricing](https://aws.amazon.com/bedrock/pricing/). A `$0.00` or wildly-off result means the table does not know the model and budgets cannot be trusted.
 6. The doc-drift test (`cdk/test/contracts/model-default-docs-parity.test.ts`) fails until the documented defaults here and in `agent/README.md` match `config.py`. That failure is the reminder, not a nuisance — update both.
+7. **Tell operators who override `bedrockModels`.** The override REPLACES the default list, and synth fails ("omits the platform default model") if it lacks either new default. A deployment carrying `-c bedrockModels=[...]` or a `cdk.json` `context` entry must add the new bare IDs before pulling the bump — call this out in the release notes.
 
 ### Cost and model selection
 
diff --git a/docs/src/content/docs/developer-guide/Repository-preparation.md b/docs/src/content/docs/developer-guide/Repository-preparation.md
index 9190be278..0cfea27ae 100644
--- a/docs/src/content/docs/developer-guide/Repository-preparation.md
+++ b/docs/src/content/docs/developer-guide/Repository-preparation.md
@@ -48,7 +48,7 @@ definitions.push({
   repo: 'acme/my-service',
   compute: { runtimeArn: '...' },                    // override the default runtime ARN
   agent: {
-    modelId: 'global.anthropic.claude-opus-5',       // foundation model override
+    modelId: 'global.anthropic.claude-opus-5-5',     // foundation model override
     maxTurns: 150,                                    // default turn limit for this repo
     systemPromptOverrides: 'Extra instructions...',   // appended to the platform prompt
   },
diff --git a/docs/src/content/docs/getting-started/Deployment-guide.md b/docs/src/content/docs/getting-started/Deployment-guide.md
index 832b21d0a..53488f40e 100644
--- a/docs/src/content/docs/getting-started/Deployment-guide.md
+++ b/docs/src/content/docs/getting-started/Deployment-guide.md
@@ -217,7 +217,7 @@ At public US East (N. Virginia) first-tier list rates verified in August 2026, t
 
 | Service | Used By | Scales to Zero |
 |---------|---------|---------------|
-| Bedrock (Claude Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5) | Agent reasoning, cross-region inference profiles | Yes |
+| Bedrock (Claude Opus 5.5, Haiku 5.5, plus Sonnet 4.6, Opus 4.8, Opus 5, Haiku 4.5 for pinned repos) | Agent reasoning, cross-region inference profiles | Yes |
 | Bedrock Guardrails | Prompt injection detection on task input | Yes |
 | Bedrock AgentCore Memory | Semantic + episodic extraction strategies | Yes |
 | AWS Agent Registry (preview, default-on) | Versioned agent asset catalog and governance | N/A (managed preview service) |