diff --git a/.changeset/canonical-format-validate-input-required-tools.md b/.changeset/canonical-format-validate-input-required-tools.md new file mode 100644 index 00000000000..766804d7109 --- /dev/null +++ b/.changeset/canonical-format-validate-input-required-tools.md @@ -0,0 +1,5 @@ +--- +"adcontextprotocol": patch +--- + +`canonical_format_validate_input` no longer lists `comply_test_controller` in `required_tools`. The storyboard runner's per-storyboard gate admits a storyboard when any listed tool is present, so any agent exposing the (universal) test controller was selected and then failed all 17 steps on the missing `validate_input`. Agents without `validate_input` now receive a coverage-gap skip; agents implementing it run unchanged. Same shape as #6774 (fixes #7404, bug 1). diff --git a/.github/workflows/training-agent-storyboards.yml b/.github/workflows/training-agent-storyboards.yml index 7e7cbb76c68..d8a5462d7ec 100644 --- a/.github/workflows/training-agent-storyboards.yml +++ b/.github/workflows/training-agent-storyboards.yml @@ -80,7 +80,7 @@ jobs: - surface: current tenant: governance min_clean_storyboards: 47 - min_passing_steps: 161 + min_passing_steps: 160 - surface: current tenant: creative min_clean_storyboards: 49 @@ -92,7 +92,7 @@ jobs: - surface: current tenant: brand min_clean_storyboards: 45 - min_passing_steps: 116 + min_passing_steps: 115 - surface: current tenant: si min_clean_storyboards: 42 diff --git a/scripts/run-storyboards-matrix.sh b/scripts/run-storyboards-matrix.sh index a826a103d91..641fe1f8548 100755 --- a/scripts/run-storyboards-matrix.sh +++ b/scripts/run-storyboards-matrix.sh @@ -269,10 +269,10 @@ else # clean-result-row and passing-step regression floors. "signals:45:80" "sales:133:632" - "governance:47:161" + "governance:47:160" "creative:49:209" "creative-builder:50:184" - "brand:45:116" + "brand:45:115" "si:42:50" ) fi diff --git a/static/compliance/source/universal/canonical-format-validate-input.yaml b/static/compliance/source/universal/canonical-format-validate-input.yaml index ab410bf422a..1d993a8c9e7 100644 --- a/static/compliance/source/universal/canonical-format-validate-input.yaml +++ b/static/compliance/source/universal/canonical-format-validate-input.yaml @@ -7,7 +7,6 @@ track: creative introduced_in: "3.1" required_tools: - validate_input - - comply_test_controller requires: - controller diff --git a/tests/run-storyboards-sharding.test.cjs b/tests/run-storyboards-sharding.test.cjs index c95a719ddf9..033ffbc6d98 100644 --- a/tests/run-storyboards-sharding.test.cjs +++ b/tests/run-storyboards-sharding.test.cjs @@ -300,13 +300,19 @@ test('creative-builder uses bounded isolated children in local and CI matrices', test('current training-agent floors are ratcheted and mirrored by local and CI runners', () => { const workflow = fs.readFileSync(STORYBOARD_WORKFLOW, 'utf8'); const matrixRunner = fs.readFileSync(MATRIX_RUNNER, 'utf8'); + // governance and brand dropped one passing step each when + // canonical_format_validate_input stopped listing comply_test_controller in + // required_tools: both tenants expose the controller but not validate_input, + // so the storyboard is no longer selected for them and the single step it + // contributed is gone. Deliberate de-ratchet, not a regression — the + // clean-storyboard floors are untouched. const baselines = [ ['signals', 45, 80], ['sales', 133, 632], - ['governance', 47, 161], + ['governance', 47, 160], ['creative', 49, 209], ['creative-builder', 50, 184], - ['brand', 45, 116], + ['brand', 45, 115], ['si', 42, 50], ];