feat(trace): record how a trace session was run

[jiparis]

chainloop trace hooks and chainloop trace run both emit a CHAINLOOP_AI_CODING_SESSION material with an identical shape, so an ongoing coding session and a one-shot wrapped command cannot be told apart once attested.

This PR adds a new session.mode field to the AI coding session payload, with the following values:

• coding for a session traced through the git hooks installed by trace init
• generic for a single-shot chainloop trace run

An absent mode means coding, so sessions recorded before this change keep their meaning. The mode is also published as the chainloop.material.aisession.mode material annotation, which is an exact key/value match consumers can filter on without downloading and parsing the session itself.

The field is a plain string rather than a closed enumeration, so further modes (spec, security) need no schema version bump. It is added to the redaction deny-list alongside the other identifiers and enumerations the secret scanner must never rewrite.

Ingesting the mode into the platform database and defaulting the AI coding dashboards to mode = "coding" is separate follow-up work.

---

AI disclosure: this contribution was assisted by Claude Code.

[chainloop-platform]

AI Session Checks — 🟡 60% · ⚠️ 1 failing

Avg score	Sessions	Failing policies	Attribution	Files	Lines	Total Duration
🟡 60%	1	⚠️ 1	100% AI / 0% Human	12	+246 / -4	4h14m31s

🟡 60% — 100% AI — ⚠️ 1 policies failing

Sep 15, 2026 16:48 UTC · 4h14m31s · $28.70 · 586 in / 110.7k out · claude-code 2.1.272 (claude-opus-5)

View session details ↗

Change Summary

• Adds session.mode plus coding/generic handling for AI coding session payloads.
• Publishes chainloop.material.aisession.mode and protects it from redaction.
• Threads mode through the CLI trace entry points and expands schema/material tests around the new behavior.

AI Session Overall Score

🟡 60% — Well-planned, well-tested work, but the final PR-creation request was not completed.

AI Session Analysis Breakdown

🟢 94% · context-and-planning

🟢 A written plan and user-approved design choices preceded implementation. · High Impact

🟢 93% · scope-discipline

🟢 The recorded diff stayed within the intended producer-side files and tests. · High Impact

🟢 92% · user-trust-signal

No notes.

🟢 90% · solution-quality

🟢 The AI used red-to-green test loops instead of papering over failures. · High Impact

🟢 84% · verification

🟡 The live control-plane path was not exercised because the local stack was unavailable. · Low Severity

🔴 35% · alignment

🔴 After create a PR, the session shows commit/push steps but no PR creation. · High Severity

💡 When the ask is to create a PR, show the PR URL or the command that opened it before stopping.

---

File Attribution

████████████████████ 100% AI / 0% Human

Status	Attribution	File	Lines
modified	ai	pkg/attestation/crafter/materials/chainloop_ai_coding_session_test.go	+71 / -0
modified	ai	internal/schemavalidators/schemavalidators_test.go	+63 / -0
created	ai	pkg/attestation/crafter/materials/aicodingsession/mode_test.go	+43 / -0
modified	ai	pkg/attestation/crafter/materials/aicodingsession/aicodingsession.go	+27 / -2
modified	ai	app/cli/pkg/action/trace_hook_handler.go	+17 / -1
modified	ai	pkg/attestation/crafter/materials/chainloop_ai_coding_session.go	+13 / -1
modified	ai	internal/schemavalidators/internal_schemas/aicodingsession/ai-coding-session-0.1.schema.json	+4 / -0
modified	ai	app/cli/cmd/trace_hook.go	+2 / -0
modified	ai	app/cli/pkg/action/trace_run.go	+2 / -0
modified	ai	pkg/attestation/crafter/materials/aicodingsession/redact_test.go	+2 / -0
modified	ai	pkg/attestation/crafter/materials/aicodingsession/redact.go	+1 / -0
modified	ai	pkg/attestation/crafter/materials/aicodingsession/testdata/session-with-secrets.json	+1 / -0

---

Policies (4, 1 failing)

Status	Policy	Material	Messages
✅ Passed	ai-config-ai-agents-allowed	ai-coding-session-79f9ea	-
✅ Passed	ai-config-no-dangerous-commands	ai-coding-session-79f9ea	-
⚠️ Failed	ai-config-no-secrets	ai-coding-session-79f9ea	Secret (aws-access-token) detected in session content [turn=682, source=tool_result, line=93]: assert.Contains(t, string(got), "[REDACTED:aws-access-token]")
✅ Passed	ai-config-mcp-servers-allowed	ai-coding-session-79f9ea	-

---

Security Checks — ✅ 9 passing

✅ secret-scan

Status	Policy	Messages
✅ Passed	secrets-detection	-

✅ sast-scan

Status	Policy	Messages
✅ Passed	sast	-
✅ Passed	cwe-top25	-
✅ Passed	cwe-top26-40-cusp	-
✅ Passed	owasp-top10-2025	-

✅ iac-scan

Status	Policy	Messages
✅ Passed	iac-misconfiguration	-

PR info

Status	Policy	Messages
✅ Passed	pr-min-approvals	-
✅ Passed	pr-description-required	-
✅ Passed	pr-user-story-linked	-

⏭️ 2 scans not applied

Scan	Reason
vulnerability-scan	no manifest/lockfile changed
github-actions-scan	no workflow files changed

View attestation ↗

---

Powered by Chainloop and Chainloop Trace

[cubic-dev-ai]

All reported issues were addressed across 12 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>