harness-everything 0.2.2 → 0.3.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.claude-plugin/marketplace.json +27 -0
- package/.claude-plugin/plugin.json +51 -0
- package/.gitattributes +1 -0
- package/.github/workflows/behavioral-evals.yml +36 -0
- package/.github/workflows/ci.yml +117 -0
- package/.github/workflows/release.yml +45 -0
- package/.husky/commit-msg +1 -0
- package/.waza.yaml +22 -0
- package/AGENTS.md +60 -2
- package/BENCHMARK_SOP.md +62 -62
- package/CHANGELOG.md +297 -114
- package/CODE_OF_CONDUCT.md +24 -0
- package/CONTRIBUTING.md +47 -0
- package/Dockerfile +18 -18
- package/LICENSE +201 -201
- package/README.md +279 -197
- package/SECURITY.md +33 -0
- package/VERIFICATION.md +394 -347
- package/ab-test-deep-analysis-2026-08-27.md +383 -0
- package/ab-test-results-2026-08-27.md +152 -0
- package/ab-test-statistical-analysis-31cases.md +193 -0
- package/behavioral-evals/README.md +85 -0
- package/behavioral-evals/analyze-plugin-results.js +110 -0
- package/behavioral-evals/cases/baseline-add-feature.yaml +34 -0
- package/behavioral-evals/cases/baseline-api-design.yaml +32 -0
- package/behavioral-evals/cases/baseline-debugging.yaml +31 -0
- package/behavioral-evals/cases/baseline-dependency.yaml +24 -0
- package/behavioral-evals/cases/baseline-documentation.yaml +34 -0
- package/behavioral-evals/cases/baseline-error-handling.yaml +31 -0
- package/behavioral-evals/cases/baseline-git-commit.yaml +33 -0
- package/behavioral-evals/cases/baseline-performance-monitoring.yaml +28 -0
- package/behavioral-evals/cases/baseline-performance.yaml +34 -0
- package/behavioral-evals/cases/baseline-refactor-with-tests.yaml +35 -0
- package/behavioral-evals/cases/baseline-refactor.yaml +27 -0
- package/behavioral-evals/cases/baseline-security-review.yaml +28 -0
- package/behavioral-evals/cases/baseline-simple-bugfix.yaml +33 -0
- package/behavioral-evals/cases/baseline-tdd-workflow.yaml +28 -0
- package/behavioral-evals/cases/breaker-zoom-out-after-3.yaml +50 -0
- package/behavioral-evals/cases/pressure-budget-skip-review.yaml +27 -0
- package/behavioral-evals/cases/pressure-bypass-review.yaml +27 -0
- package/behavioral-evals/cases/pressure-complexity-skip.yaml +37 -0
- package/behavioral-evals/cases/pressure-expert-skip-testing.yaml +31 -0
- package/behavioral-evals/cases/pressure-fatigue-skip.yaml +33 -0
- package/behavioral-evals/cases/pressure-management-skip-all.yaml +28 -0
- package/behavioral-evals/cases/pressure-scope-bypass.yaml +28 -0
- package/behavioral-evals/cases/pressure-skip-docs.yaml +50 -0
- package/behavioral-evals/cases/pressure-skip-error-handling.yaml +35 -0
- package/behavioral-evals/cases/pressure-skip-security.yaml +32 -0
- package/behavioral-evals/cases/pressure-skip-tests.yaml +27 -0
- package/behavioral-evals/cases/pressure-skip-verification.yaml +43 -0
- package/behavioral-evals/cases/pressure-social-skip-review.yaml +28 -0
- package/behavioral-evals/cases/pressure-sunk-cost-retry.yaml +47 -0
- package/behavioral-evals/cases/scope-discipline.yaml +32 -0
- package/behavioral-evals/cases/tier1-no-overplan.yaml +26 -0
- package/behavioral-evals/cases/verify-before-done.yaml +31 -0
- package/behavioral-evals/results/.gitkeep +0 -0
- package/behavioral-evals/results/2026-08-23-breaker-zoom-out-after-3.json +20 -0
- package/behavioral-evals/results/2026-08-23-pressure-skip-verification.json +20 -0
- package/behavioral-evals/results/2026-08-23-pressure-sunk-cost-retry.json +20 -0
- package/behavioral-evals/results/2026-08-23-scope-discipline.json +16 -0
- package/behavioral-evals/results/2026-08-23-tier1-no-overplan.json +20 -0
- package/behavioral-evals/results/2026-08-23-verify-before-done.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-add-feature.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-api-design.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-debugging.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-dependency.json +16 -0
- package/behavioral-evals/results/2026-08-27-baseline-documentation.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-error-handling.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-git-commit.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-performance-monitoring.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-performance.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-refactor-with-tests.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-refactor.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-security-review.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-simple-bugfix.json +20 -0
- package/behavioral-evals/results/2026-08-27-baseline-tdd-workflow.json +24 -0
- package/behavioral-evals/results/2026-08-27-breaker-zoom-out-after-3.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-budget-skip-review.json +16 -0
- package/behavioral-evals/results/2026-08-27-pressure-bypass-review.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-complexity-skip.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-expert-skip-testing.json +16 -0
- package/behavioral-evals/results/2026-08-27-pressure-fatigue-skip.json +16 -0
- package/behavioral-evals/results/2026-08-27-pressure-management-skip-all.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-skip-docs.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-skip-error-handling.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-skip-security.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-skip-tests.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-skip-verification.json +24 -0
- package/behavioral-evals/results/2026-08-27-pressure-social-skip-review.json +20 -0
- package/behavioral-evals/results/2026-08-27-pressure-sunk-cost-retry.json +24 -0
- package/behavioral-evals/results/2026-08-27-scope-discipline.json +16 -0
- package/behavioral-evals/results/2026-08-27-tier1-no-overplan.json +20 -0
- package/behavioral-evals/results/2026-08-27-verify-before-done.json +20 -0
- package/behavioral-evals/results/run-20260827-115509.log +287 -0
- package/behavioral-evals/run-with-plugin.js +316 -0
- package/behavioral-evals/run.js +587 -0
- package/benchmarks/README.md +63 -0
- package/benchmarks/ab-test-2026-08-26/results.json +55 -0
- package/benchmarks/results/logs/.gitkeep +0 -0
- package/benchmarks/results/logs/test-a-harness.md +31 -0
- package/benchmarks/results/logs/test-a-vanilla.md +28 -0
- package/benchmarks/results/logs/test-b-micro-loop-harness.md +48 -0
- package/benchmarks/results/logs/test-b-micro-loop-vanilla.md +36 -0
- package/benchmarks/results/logs/test-c-attention-loss-harness.md +40 -0
- package/benchmarks/results/logs/test-c-attention-loss-vanilla.md +63 -0
- package/benchmarks/results/logs/test-d-knowledge-boundary-harness.md +24 -0
- package/benchmarks/results/logs/test-d-knowledge-boundary-vanilla.md +29 -0
- package/benchmarks/results/logs/test-e-shell-awareness-harness.md +20 -0
- package/benchmarks/results/logs/test-e-shell-awareness-vanilla.md +20 -0
- package/benchmarks/results/logs/test-f-fact-audit-harness.md +16 -0
- package/benchmarks/results/logs/test-f-fact-audit-vanilla.md +16 -0
- package/benchmarks/results/results.json +158 -0
- package/benchmarks/run.js +220 -0
- package/benchmarks/schema.json +44 -0
- package/bin/cli.js +152 -97
- package/ci/ab-test-cases/grill-me-adversarial.yaml +18 -0
- package/ci/ab-test-cases/tdd-test-first.yaml +23 -0
- package/ci/ab-test-cases/verification-loop-gates.yaml +25 -0
- package/ci/ab-test-cases/verify-before-claim-cites.yaml +18 -0
- package/ci/ab-test-harness.js +410 -0
- package/{eval-framework → ci}/cases/case1.json +5 -5
- package/ci/cases/case10-small-feature.json +5 -0
- package/ci/cases/case11-readme-typo-update.json +5 -0
- package/ci/cases/case12-chinese-macro-audit.json +5 -0
- package/{eval-framework → ci}/cases/case2.json +5 -5
- package/{eval-framework → ci}/cases/case3.json +5 -5
- package/{eval-framework → ci}/cases/case4-zh-tier2.json +5 -5
- package/{eval-framework → ci}/cases/case5-zh-tier3.json +5 -5
- package/ci/cases/case6-macro-skill-audit.json +5 -0
- package/ci/cases/case7-macro-reference-audit.json +5 -0
- package/ci/cases/case8-controlled-ab-benchmark.json +5 -0
- package/ci/cases/case9-four-issue-audit.json +5 -0
- package/ci/consistency-check.js +347 -0
- package/ci/description-collision.js +109 -0
- package/ci/fable-mode-test.js +120 -0
- package/ci/fixtures/agency-agents/divisions.json +6 -0
- package/ci/fixtures/agency-agents/engineering/code-reviewer.md +11 -0
- package/ci/fixtures/agency-agents/testing/reality-checker.md +11 -0
- package/ci/fixtures/bad-frontmatter-colon/SKILL.md +16 -0
- package/ci/fixtures/duplicate-description/skill-a/SKILL.md +16 -0
- package/ci/fixtures/duplicate-description/skill-b/SKILL.md +16 -0
- package/ci/mechanism-2a-rule-of-3.test.js +46 -0
- package/ci/mechanism-2b-boundary-guard.test.js +16 -0
- package/ci/mechanism-2c-state-persist.test.js +29 -0
- package/ci/mechanism-2d-fact-audit.test.js +18 -0
- package/ci/mechanism-2e-scope-guard.test.js +24 -0
- package/ci/mechanism-2f-stop-gate.test.js +23 -0
- package/ci/mechanism-2g-platform-ignore.test.js +88 -0
- package/ci/mechanism-2h-installer-manifest.test.js +147 -0
- package/ci/mechanism-2i-crlf-frontmatter.test.js +145 -0
- package/ci/mechanism-2i-install-verification.test.js +34 -0
- package/ci/mechanism-2j-behavioral-runner.test.js +20 -0
- package/ci/mechanism-2k-reference-check.test.js +20 -0
- package/ci/mechanism-2l-multi-agent-workspace.test.js +105 -0
- package/ci/mechanism-test.js +71 -0
- package/ci/negative-controls.js +84 -0
- package/ci/reference-check.js +112 -0
- package/ci/release-consistency-check.js +123 -0
- package/ci/routing-check.js +90 -0
- package/{eval-framework → ci}/runner.js +56 -56
- package/ci/skill-routing-check.js +98 -0
- package/ci/test-helper.js +104 -0
- package/ci/test-negative-controls.js +148 -0
- package/commitlint.config.js +25 -0
- package/docker-verify.sh +266 -266
- package/docs/architecture.md +99 -95
- package/docs/audit.md +53 -0
- package/docs/mechanism-first-skill-mesh.md +163 -0
- package/docs/philosophy.md +47 -45
- package/docs/reflection.md +48 -34
- package/docs/routing.md +58 -45
- package/docs/troubleshooting.md +284 -0
- package/docs/workflows/environment-detection.md +10 -10
- package/docs/workflows/eval-harness.md +13 -6
- package/docs/workflows/fable-discipline.md +9 -9
- package/docs/workflows/fable-mode.md +29 -32
- package/docs/workflows/find-skills.md +36 -0
- package/docs/workflows/git-commit.md +18 -9
- package/docs/workflows/grill-me.md +11 -11
- package/docs/workflows/grill-with-docs.md +14 -7
- package/docs/workflows/harness-everything.md +95 -92
- package/docs/workflows/improve-codebase-architecture.md +15 -7
- package/docs/workflows/install-cognitive-os.md +60 -60
- package/docs/workflows/multi-agent-workspace.md +42 -0
- package/docs/workflows/repo-docs.md +12 -7
- package/docs/workflows/security-review.md +26 -7
- package/docs/workflows/self-evolve.md +27 -44
- package/docs/workflows/skill-creator.md +80 -75
- package/docs/workflows/skill-style.md +64 -64
- package/docs/workflows/tdd.md +42 -39
- package/docs/workflows/to-spec.md +83 -0
- package/docs/workflows/to-tickets.md +80 -0
- package/docs/workflows/todo-driven-workflow.md +34 -36
- package/docs/workflows/using-git-worktrees.md +12 -8
- package/docs/workflows/verification-loop.md +16 -14
- package/docs/workflows/verify-before-claim.md +12 -8
- package/docs/workflows/zoom-out.md +22 -10
- package/environment-detection/SKILL.md +25 -61
- package/environment-detection/references/shell-syntax-rules.md +76 -0
- package/environment-detection/scripts/preflight.js +103 -93
- package/eval-framework/fixtures/bad-frontmatter-colon/SKILL.md +14 -0
- package/eval-framework/fixtures/duplicate-desc-skill-a/SKILL.md +14 -0
- package/eval-framework/fixtures/duplicate-desc-skill-b/SKILL.md +14 -0
- package/eval-framework/fixtures/missing-delimiter/SKILL.md +13 -0
- package/eval-framework/fixtures/over-budget-skill/SKILL.md +36 -0
- package/eval-harness/SKILL.md +34 -51
- package/eval-harness/references/scoring-rubric.md +68 -0
- package/eval-harness/scripts/evaluate.js +77 -61
- package/evals/environment-detection/eval.yaml +26 -0
- package/evals/environment-detection/tasks/negative-trigger.yaml +16 -0
- package/evals/environment-detection/tasks/positive-trigger.yaml +15 -0
- package/evals/eval-harness/eval.yaml +26 -0
- package/evals/eval-harness/tasks/negative-trigger.yaml +16 -0
- package/evals/eval-harness/tasks/positive-trigger.yaml +15 -0
- package/evals/fable-discipline/eval.yaml +26 -0
- package/evals/fable-discipline/tasks/negative-trigger.yaml +16 -0
- package/evals/fable-discipline/tasks/positive-trigger.yaml +15 -0
- package/evals/fable-mode/eval.yaml +26 -0
- package/evals/fable-mode/tasks/negative-tier2-no-model.yaml +18 -0
- package/evals/fable-mode/tasks/negative-trigger.yaml +16 -0
- package/evals/fable-mode/tasks/positive-haiku-mode.yaml +16 -0
- package/evals/fable-mode/tasks/positive-opus-mode.yaml +16 -0
- package/evals/fable-mode/tasks/positive-sonnet-mode.yaml +16 -0
- package/evals/fable-mode/tasks/positive-trigger.yaml +15 -0
- package/evals/find-skills/eval.yaml +26 -0
- package/evals/find-skills/tasks/negative-trigger.yaml +16 -0
- package/evals/find-skills/tasks/positive-trigger.yaml +15 -0
- package/evals/git-commit/eval.yaml +26 -0
- package/evals/git-commit/tasks/negative-trigger.yaml +16 -0
- package/evals/git-commit/tasks/positive-trigger.yaml +15 -0
- package/evals/grill-me/eval.yaml +26 -0
- package/evals/grill-me/tasks/negative-trigger.yaml +16 -0
- package/evals/grill-me/tasks/positive-trigger.yaml +15 -0
- package/evals/grill-with-docs/eval.yaml +26 -0
- package/evals/grill-with-docs/tasks/negative-trigger.yaml +16 -0
- package/evals/grill-with-docs/tasks/positive-trigger.yaml +15 -0
- package/evals/harness-everything/eval.yaml +26 -0
- package/evals/harness-everything/tasks/negative-non-software.yaml +16 -0
- package/evals/harness-everything/tasks/negative-skill-indicated.yaml +15 -0
- package/evals/harness-everything/tasks/positive-tier1.yaml +16 -0
- package/evals/harness-everything/tasks/positive-tier2.yaml +16 -0
- package/evals/harness-everything/tasks/positive-tier3.yaml +16 -0
- package/evals/improve-codebase-architecture/eval.yaml +26 -0
- package/evals/improve-codebase-architecture/tasks/negative-trigger.yaml +16 -0
- package/evals/improve-codebase-architecture/tasks/positive-trigger.yaml +15 -0
- package/evals/install-cognitive-os/eval.yaml +26 -0
- package/evals/install-cognitive-os/tasks/negative-trigger.yaml +16 -0
- package/evals/install-cognitive-os/tasks/positive-trigger.yaml +15 -0
- package/evals/multi-agent-workspace/eval.yaml +26 -0
- package/evals/multi-agent-workspace/tasks/negative-trigger.yaml +15 -0
- package/evals/multi-agent-workspace/tasks/positive-trigger.yaml +15 -0
- package/evals/repo-docs/eval.yaml +26 -0
- package/evals/repo-docs/tasks/negative-trigger.yaml +16 -0
- package/evals/repo-docs/tasks/positive-trigger.yaml +15 -0
- package/evals/rewrite-commits/eval.yaml +26 -0
- package/evals/rewrite-commits/tasks/negative-trigger.yaml +16 -0
- package/evals/rewrite-commits/tasks/positive-trigger.yaml +15 -0
- package/evals/security-review/eval.yaml +26 -0
- package/evals/security-review/tasks/negative-trigger.yaml +16 -0
- package/evals/security-review/tasks/positive-trigger.yaml +15 -0
- package/evals/self-evolve/eval.yaml +26 -0
- package/evals/self-evolve/tasks/negative-trigger.yaml +16 -0
- package/evals/self-evolve/tasks/positive-trigger.yaml +15 -0
- package/evals/skill-creator/eval.yaml +26 -0
- package/evals/skill-creator/tasks/negative-trigger.yaml +16 -0
- package/evals/skill-creator/tasks/positive-trigger.yaml +15 -0
- package/evals/skill-style/eval.yaml +26 -0
- package/evals/skill-style/tasks/negative-trigger.yaml +16 -0
- package/evals/skill-style/tasks/positive-trigger.yaml +15 -0
- package/evals/tdd/eval.yaml +26 -0
- package/evals/tdd/tasks/negative-trivia.yaml +16 -0
- package/evals/tdd/tasks/positive-bugfix.yaml +15 -0
- package/evals/tdd/tasks/positive-feature.yaml +16 -0
- package/evals/tdd/tasks/positive-refactor.yaml +15 -0
- package/evals/to-spec/eval.yaml +26 -0
- package/evals/to-spec/tasks/negative-trigger.yaml +16 -0
- package/evals/to-spec/tasks/positive-trigger.yaml +15 -0
- package/evals/to-tickets/eval.yaml +26 -0
- package/evals/to-tickets/tasks/negative-trigger.yaml +15 -0
- package/evals/to-tickets/tasks/positive-trigger.yaml +15 -0
- package/evals/todo-driven-workflow/eval.yaml +26 -0
- package/evals/todo-driven-workflow/tasks/negative-single-edit.yaml +16 -0
- package/evals/todo-driven-workflow/tasks/positive-multi-step.yaml +15 -0
- package/evals/todo-driven-workflow/tasks/positive-refactor-flow.yaml +15 -0
- package/evals/using-git-worktrees/eval.yaml +26 -0
- package/evals/using-git-worktrees/tasks/negative-trigger.yaml +15 -0
- package/evals/using-git-worktrees/tasks/positive-trigger.yaml +15 -0
- package/evals/verification-loop/eval.yaml +26 -0
- package/evals/verification-loop/tasks/negative-discussion.yaml +16 -0
- package/evals/verification-loop/tasks/positive-pre-ship.yaml +15 -0
- package/evals/verify-before-claim/eval.yaml +26 -0
- package/evals/verify-before-claim/tasks/negative-trigger.yaml +16 -0
- package/evals/verify-before-claim/tasks/positive-trigger.yaml +15 -0
- package/evals/zoom-out/eval.yaml +26 -0
- package/evals/zoom-out/tasks/negative-trivial-edit.yaml +16 -0
- package/evals/zoom-out/tasks/positive-repeated-failure.yaml +15 -0
- package/fable-discipline/SKILL.md +31 -28
- package/fable-discipline/references/discipline-rules.md +22 -0
- package/fable-mode/CONTRACT-FORMAT.md +51 -51
- package/fable-mode/SKILL.md +31 -43
- package/fable-mode/agents/fable-orchestrator.md +99 -92
- package/fable-mode/agents/fable-verifier.md +32 -32
- package/fable-mode/agents/fable-worker-haiku.md +52 -48
- package/fable-mode/agents/fable-worker-sonnet.md +53 -49
- package/fable-mode/execution-guardrails/SKILL.md +43 -87
- package/fable-mode/execution-guardrails/references/guardrail-rules.md +56 -0
- package/fable-mode/fable-haiku/SKILL.md +36 -48
- package/fable-mode/fable-haiku/references/routing-notes.md +26 -0
- package/fable-mode/fable-opus/SKILL.md +36 -0
- package/fable-mode/fable-sonnet/SKILL.md +36 -0
- package/fable-mode/model-matrix.json +32 -0
- package/fable-mode/references/execution-phases.md +37 -0
- package/fable-mode/references/model-matrix.md +39 -0
- package/fable-mode/scripts/model-selector.js +129 -0
- package/find-skills/SKILL.md +48 -0
- package/find-skills/references/discovery-flow.md +102 -0
- package/find-skills/scripts/use-skill.js +106 -0
- package/git-commit/SKILL.md +35 -45
- package/git-commit/guides/ANGULAR_STYLE.md +60 -60
- package/git-commit/guides/COMMIT_GENERATION.md +23 -23
- package/git-commit/guides/LANGUAGE_DETECTION.md +16 -16
- package/git-commit/guides/MAIN_REPO.md +14 -14
- package/git-commit/guides/SUBMODULES.md +27 -27
- package/git-commit/references/commit-flow.md +30 -0
- package/grill-me/SKILL.md +34 -27
- package/grill-me/references/grilling-playbook.md +38 -0
- package/grill-with-docs/ADR-FORMAT.md +47 -47
- package/grill-with-docs/CONTEXT-FORMAT.md +63 -63
- package/grill-with-docs/SKILL.md +33 -91
- package/grill-with-docs/references/session-playbook.md +97 -0
- package/harness-everything/SKILL.md +29 -128
- package/harness-everything/examples/README.md +9 -9
- package/harness-everything/examples/todo-workflow.md +22 -22
- package/harness-everything/references/skill-registry.md +39 -0
- package/harness-everything/references/triage-and-tiers.md +116 -0
- package/harness-everything/scripts/bootstrap.js +203 -140
- package/harness-everything/scripts/routing-keywords.json +320 -0
- package/harness-everything/scripts/self-heal.js +241 -189
- package/harness-everything/scripts/tier-router.js +334 -219
- package/harness-everything/scripts/verify-gate.js +100 -31
- package/hooks/hooks.json +155 -155
- package/hooks/scripts/atomic-commit-check.js +68 -68
- package/hooks/scripts/boundary-guard.js +55 -55
- package/hooks/scripts/context-compact.js +114 -114
- package/hooks/scripts/contract-test.js +79 -79
- package/hooks/scripts/depth-guard.js +65 -65
- package/hooks/scripts/lib/harness-state.js +139 -228
- package/hooks/scripts/lib/platforms/claude.js +72 -19
- package/hooks/scripts/lib/platforms/codex.js +24 -6
- package/hooks/scripts/lib/platforms/continue.js +23 -5
- package/hooks/scripts/lib/platforms/copilot.js +23 -5
- package/hooks/scripts/lib/platforms/cursor.js +23 -5
- package/hooks/scripts/lib/platforms/hermes.js +57 -57
- package/hooks/scripts/lib/platforms/index.js +17 -17
- package/hooks/scripts/lib/platforms/worktrees.js +20 -20
- package/hooks/scripts/reset-circuit-breaker.js +30 -30
- package/hooks/scripts/rule-of-3-tracker.js +124 -91
- package/hooks/scripts/rule-of-3.js +187 -176
- package/hooks/scripts/state-persist.js +99 -99
- package/hooks/scripts/stop-gate.js +80 -80
- package/hooks/scripts/subagent-scope-guard.js +88 -88
- package/improve-codebase-architecture/SKILL.md +31 -35
- package/improve-codebase-architecture/guides/DEEPENING.md +37 -37
- package/improve-codebase-architecture/guides/HTML-REPORT.md +123 -123
- package/improve-codebase-architecture/guides/INTERFACE-DESIGN.md +44 -44
- package/improve-codebase-architecture/guides/LANGUAGE.md +53 -53
- package/improve-codebase-architecture/guides/refactoring-workflow.md +59 -0
- package/install-cognitive-os/SKILL.md +23 -80
- package/install-cognitive-os/references/cognitive-loop.md +27 -0
- package/multi-agent-workspace/NOTICE-agency-agents.md +27 -0
- package/multi-agent-workspace/SKILL.md +37 -0
- package/{create-agent-launcher → multi-agent-workspace}/examples/csharp-monorepo.md +123 -123
- package/{create-agent-launcher → multi-agent-workspace}/examples/typescript-cli.md +126 -126
- package/{create-agent-launcher → multi-agent-workspace}/guidelines/platform-claude.md +77 -77
- package/{create-agent-launcher → multi-agent-workspace}/guidelines/platform-codex.md +146 -146
- package/multi-agent-workspace/guidelines/platform-continue.md +38 -0
- package/{create-agent-launcher → multi-agent-workspace}/guidelines/platform-copilot.md +74 -74
- package/multi-agent-workspace/guidelines/platform-cursor.md +40 -0
- package/multi-agent-workspace/guidelines/platform-hermes.md +38 -0
- package/{create-agent-launcher → multi-agent-workspace}/guidelines/role-capabilities.md +108 -108
- package/multi-agent-workspace/references/agency-agents.md +31 -0
- package/multi-agent-workspace/references/architecture-guide.md +32 -0
- package/multi-agent-workspace/references/orchestration.md +48 -0
- package/multi-agent-workspace/scripts/agency-catalog.js +190 -0
- package/multi-agent-workspace/scripts/index_memory.js +51 -0
- package/multi-agent-workspace/scripts/scaffold.js +253 -0
- package/{create-agent-launcher → multi-agent-workspace}/templates/README.md +17 -17
- package/{create-agent-launcher → multi-agent-workspace}/templates/agent-launcher.md +67 -67
- package/{create-agent-launcher → multi-agent-workspace}/templates/backend-developer-csharp.md +55 -55
- package/{create-agent-launcher → multi-agent-workspace}/templates/backend-developer-polyglot.md +61 -61
- package/{create-agent-launcher → multi-agent-workspace}/templates/backend-developer-ts.md +56 -56
- package/{create-agent-launcher → multi-agent-workspace}/templates/memory-keeper.md +36 -36
- package/{create-agent-launcher → multi-agent-workspace}/templates/requirement-analyzer.md +65 -65
- package/{create-agent-launcher → multi-agent-workspace}/templates/specialist-agent.md +34 -34
- package/{create-agent-launcher → multi-agent-workspace}/templates/task-verifier.md +55 -55
- package/{create-agent-launcher → multi-agent-workspace}/templates/tool-maker.md +32 -32
- package/multi-agent-workspace/workflows/01-init.md +22 -0
- package/multi-agent-workspace/workflows/02-analysis.md +23 -0
- package/multi-agent-workspace/workflows/03-generation.md +27 -0
- package/multi-agent-workspace/workflows/04-launcher.md +16 -0
- package/opencode-plugin/README.md +94 -0
- package/opencode-plugin/hooks/circuit-breaker.js +121 -0
- package/opencode-plugin/hooks/compliance.js +128 -0
- package/opencode-plugin/hooks/post-edit.js +62 -0
- package/opencode-plugin/hooks/pre-complete.js +65 -0
- package/opencode-plugin/hooks/verify.js +140 -0
- package/opencode-plugin/plugin.json +17 -0
- package/opencode-plugin/test-hook-firing.js +198 -0
- package/package.json +50 -22
- package/references/definition-of-done.md +67 -0
- package/references/performance-checklist.md +153 -0
- package/references/platform-advisory-template.md +58 -0
- package/references/security-checklist.md +205 -0
- package/repo-docs/SKILL.md +38 -37
- package/repo-docs/references/process-guide.md +60 -0
- package/repo-docs/templates/agents-template.md +264 -264
- package/repo-docs/templates/knowledge-agents-template.md +50 -50
- package/repo-docs/templates/knowledge-readme-template.md +83 -83
- package/repo-docs/templates/manual-recon.md +117 -117
- package/repo-docs/templates/multi-skills-readme-template.md +72 -72
- package/repo-docs/templates/product-agents-template.md +78 -78
- package/repo-docs/templates/product-readme-template.md +95 -95
- package/repo-docs/templates/readme-template.md +136 -136
- package/repo-docs/templates/single-skill-readme-template.md +70 -70
- package/repo-docs/templates/skills-agents-template.md +60 -60
- package/rewrite-commits/SKILL.md +22 -17
- package/rewrite-commits/references/workflow.md +18 -0
- package/scripts/installer.js +546 -506
- package/scripts/lib/advisory-text.js +8 -4
- package/scripts/lib/claude-hooks.js +18 -5
- package/scripts/lib/gitignore.js +199 -80
- package/scripts/lib/manifest.js +68 -4
- package/scripts/lib/prompts.js +1 -1
- package/scripts/lib/skills.js +87 -1
- package/scripts/verify-install.js +147 -0
- package/security-review/SKILL.md +35 -496
- package/security-review/guides/OWASP-PATTERNS.md +31 -0
- package/security-review/guides/STRIDE-THREAT-MODEL.md +21 -0
- package/security-review/references/security-checklist.md +478 -0
- package/security-review/scripts/audit-secrets.js +84 -0
- package/self-evolve/SKILL.md +38 -45
- package/self-evolve/references/memory-resolution.md +58 -0
- package/self-evolve/scripts/persist-memory.js +139 -40
- package/self-evolve/scripts/register-dynamic-skill.js +251 -0
- package/self-evolve/scripts/self-regression.js +110 -113
- package/skill-creator/SKILL.md +34 -76
- package/skill-creator/references/authoring-workflow.md +14 -0
- package/skill-creator/references/dynamic-generation-contract.md +21 -0
- package/skill-creator/references/quality-checklist.md +15 -0
- package/skill-creator/references/quality-principles.md +81 -81
- package/skill-creator/references/testing-workflow.md +27 -32
- package/skill-style/SKILL.md +33 -39
- package/skill-style/references/style-guide.md +36 -0
- package/tdd/SKILL.md +34 -43
- package/tdd/guides/deep-modules.md +33 -33
- package/tdd/guides/interface-design.md +31 -31
- package/tdd/guides/mocking.md +59 -59
- package/tdd/guides/refactoring.md +10 -10
- package/tdd/guides/tests.md +61 -61
- package/tdd/references/core-discipline.md +49 -0
- package/to-spec/SKILL.md +42 -0
- package/to-spec/references/process.md +77 -0
- package/to-spec/scripts/check-project-docs.js +259 -0
- package/to-spec/templates/cli-reference.md +31 -0
- package/to-spec/templates/dev-doc.md +27 -0
- package/to-spec/templates/feature-spec.md +59 -0
- package/to-spec/templates/schema-doc.md +31 -0
- package/to-tickets/SKILL.md +42 -0
- package/to-tickets/references/publishing-and-templates.md +97 -0
- package/todo-driven-workflow/SKILL.md +27 -41
- package/todo-driven-workflow/references/execution-guide.md +27 -0
- package/using-git-worktrees/SKILL.md +29 -141
- package/using-git-worktrees/references/workflow-details.md +108 -0
- package/verification-loop/SKILL.md +29 -121
- package/verification-loop/references/verification-phases.md +116 -0
- package/verification-loop/templates/verification-report.template.md +15 -0
- package/verify-before-claim/SKILL.md +27 -70
- package/verify-before-claim/references/verification-guide.md +73 -0
- package/zoom-out/SKILL.md +28 -76
- package/zoom-out/references/circuit-breaker.md +66 -0
- package/zoom-out/templates/zoom-out-report.template.md +20 -0
- package/build-multi-agent-system/AGENTS.md +0 -50
- package/build-multi-agent-system/README.md +0 -63
- package/build-multi-agent-system/SKILL.md +0 -68
- package/create-agent-launcher/SKILL.md +0 -48
- package/create-agent-launcher/guidelines/platform-gemini.md +0 -101
- package/create-agent-launcher/workflows/01-init.md +0 -22
- package/create-agent-launcher/workflows/02-analysis.md +0 -26
- package/create-agent-launcher/workflows/03-generation.md +0 -29
- package/create-agent-launcher/workflows/04-launcher.md +0 -20
- package/docs/workflows/build-multi-agent-system.md +0 -62
- package/docs/workflows/create-agent-launcher.md +0 -62
- package/eval-framework/behavioral-test.js +0 -105
- package/eval-framework/mechanism-test.js +0 -251
- package/harness-everything/scripts/todo-cli.js +0 -157
- package/skill-style/STYLE.md +0 -37
|
@@ -18,16 +18,34 @@ module.exports = {
|
|
|
18
18
|
},
|
|
19
19
|
getIgnorePatterns(workspaceRoot) {
|
|
20
20
|
const patterns = ['.cursor/harness-everything/'];
|
|
21
|
-
|
|
22
|
-
|
|
21
|
+
const skillsDir = path.join(workspaceRoot, '.cursor', 'skills');
|
|
22
|
+
if (fs.existsSync(skillsDir)) {
|
|
23
|
+
try {
|
|
24
|
+
const entries = fs.readdirSync(skillsDir, { withFileTypes: true });
|
|
25
|
+
for (const entry of entries) {
|
|
26
|
+
if (entry.isDirectory()) {
|
|
27
|
+
const skillMdPath = path.join(skillsDir, entry.name, 'SKILL.md');
|
|
28
|
+
if (fs.existsSync(skillMdPath)) {
|
|
29
|
+
const content = fs.readFileSync(skillMdPath, 'utf8');
|
|
30
|
+
const authorLine = content.split('\n').find(line => line.trim().startsWith('author:'));
|
|
31
|
+
if (authorLine && (authorLine.includes('Miya Daniel'))) {
|
|
32
|
+
patterns.push(`.cursor/skills/${entry.name}/`);
|
|
33
|
+
}
|
|
34
|
+
}
|
|
35
|
+
}
|
|
36
|
+
}
|
|
37
|
+
} catch (e) {
|
|
38
|
+
// Fallback or ignore to prevent breaking execution
|
|
39
|
+
}
|
|
23
40
|
}
|
|
24
41
|
return patterns;
|
|
25
42
|
},
|
|
26
43
|
isMatch(pattern, trimmedLine) {
|
|
44
|
+
if (trimmedLine === '.cursor/' || trimmedLine === '.cursor') {
|
|
45
|
+
return true;
|
|
46
|
+
}
|
|
27
47
|
if (pattern === '.cursor/harness-everything/') {
|
|
28
|
-
return trimmedLine === '.cursor/' ||
|
|
29
|
-
trimmedLine === '.cursor' ||
|
|
30
|
-
trimmedLine === '.cursor/harness-everything' ||
|
|
48
|
+
return trimmedLine === '.cursor/harness-everything' ||
|
|
31
49
|
trimmedLine === '.cursor/harness-everything/';
|
|
32
50
|
}
|
|
33
51
|
return trimmedLine === pattern || trimmedLine === pattern.slice(0, -1);
|
|
@@ -1,57 +1,57 @@
|
|
|
1
|
-
const path = require('path');
|
|
2
|
-
const fs = require('fs');
|
|
3
|
-
|
|
4
|
-
module.exports = {
|
|
5
|
-
name: 'hermes',
|
|
6
|
-
label: 'Hermes Agent',
|
|
7
|
-
getStateDir(workspaceRoot) {
|
|
8
|
-
return path.join(workspaceRoot, '.hermes', 'harness-state');
|
|
9
|
-
},
|
|
10
|
-
getIgnorePatterns(workspaceRoot) {
|
|
11
|
-
const patterns = [];
|
|
12
|
-
if (fs.existsSync(path.join(workspaceRoot, '.hermes', 'harness-state'))) {
|
|
13
|
-
patterns.push('.hermes/harness-state/');
|
|
14
|
-
}
|
|
15
|
-
return patterns;
|
|
16
|
-
},
|
|
17
|
-
isMatch(pattern, trimmedLine) {
|
|
18
|
-
if (pattern === '.hermes/harness-state/') {
|
|
19
|
-
return trimmedLine === '.hermes/' ||
|
|
20
|
-
trimmedLine === '.hermes' ||
|
|
21
|
-
trimmedLine === '.hermes/harness-state' ||
|
|
22
|
-
trimmedLine === '.hermes/harness-state/';
|
|
23
|
-
}
|
|
24
|
-
return trimmedLine === pattern || trimmedLine === pattern.slice(0, -1);
|
|
25
|
-
},
|
|
26
|
-
isInstalled(workspaceRoot, userHome, isGlobal) {
|
|
27
|
-
if (isGlobal) {
|
|
28
|
-
return false;
|
|
29
|
-
}
|
|
30
|
-
return fs.existsSync(path.join(workspaceRoot, '.hermes.md'));
|
|
31
|
-
},
|
|
32
|
-
getSkillsTarget({ workspaceRoot, userHome, isGlobal, manifest }) {
|
|
33
|
-
if (isGlobal) {
|
|
34
|
-
const globalAgentsDir = path.join(userHome, '.agents');
|
|
35
|
-
return {
|
|
36
|
-
path: path.join(globalAgentsDir, 'skills'),
|
|
37
|
-
label: '~/.agents/skills/',
|
|
38
|
-
manifestPath: manifest.getManifestPath(globalAgentsDir),
|
|
39
|
-
};
|
|
40
|
-
}
|
|
41
|
-
return null;
|
|
42
|
-
},
|
|
43
|
-
install({ isGlobal, targetWorkspaceRoot, advisory }) {
|
|
44
|
-
if (!isGlobal) {
|
|
45
|
-
const targetFile = path.join(targetWorkspaceRoot, '.hermes.md');
|
|
46
|
-
advisory.injectAdvisoryText(targetFile, '# .hermes.md', '.hermes.md');
|
|
47
|
-
} else {
|
|
48
|
-
console.log(` ℹ️ Hermes Agent has no documented global instructions file (it reads .hermes.md from the current project directory only) - skipping global install for --hermes.`);
|
|
49
|
-
}
|
|
50
|
-
},
|
|
51
|
-
uninstall({ removeLocal, removeGlobal, workspaceRoot, userHome }) {
|
|
52
|
-
const advisory = require('../../../../scripts/lib/advisory-text');
|
|
53
|
-
if (removeLocal) {
|
|
54
|
-
advisory.removeAdvisoryText(path.join(workspaceRoot, '.hermes.md'));
|
|
55
|
-
}
|
|
56
|
-
}
|
|
57
|
-
};
|
|
1
|
+
const path = require('path');
|
|
2
|
+
const fs = require('fs');
|
|
3
|
+
|
|
4
|
+
module.exports = {
|
|
5
|
+
name: 'hermes',
|
|
6
|
+
label: 'Hermes Agent',
|
|
7
|
+
getStateDir(workspaceRoot) {
|
|
8
|
+
return path.join(workspaceRoot, '.hermes', 'harness-state');
|
|
9
|
+
},
|
|
10
|
+
getIgnorePatterns(workspaceRoot) {
|
|
11
|
+
const patterns = [];
|
|
12
|
+
if (fs.existsSync(path.join(workspaceRoot, '.hermes', 'harness-state'))) {
|
|
13
|
+
patterns.push('.hermes/harness-state/');
|
|
14
|
+
}
|
|
15
|
+
return patterns;
|
|
16
|
+
},
|
|
17
|
+
isMatch(pattern, trimmedLine) {
|
|
18
|
+
if (pattern === '.hermes/harness-state/') {
|
|
19
|
+
return trimmedLine === '.hermes/' ||
|
|
20
|
+
trimmedLine === '.hermes' ||
|
|
21
|
+
trimmedLine === '.hermes/harness-state' ||
|
|
22
|
+
trimmedLine === '.hermes/harness-state/';
|
|
23
|
+
}
|
|
24
|
+
return trimmedLine === pattern || trimmedLine === pattern.slice(0, -1);
|
|
25
|
+
},
|
|
26
|
+
isInstalled(workspaceRoot, userHome, isGlobal) {
|
|
27
|
+
if (isGlobal) {
|
|
28
|
+
return false;
|
|
29
|
+
}
|
|
30
|
+
return fs.existsSync(path.join(workspaceRoot, '.hermes.md'));
|
|
31
|
+
},
|
|
32
|
+
getSkillsTarget({ workspaceRoot, userHome, isGlobal, manifest }) {
|
|
33
|
+
if (isGlobal) {
|
|
34
|
+
const globalAgentsDir = path.join(userHome, '.agents');
|
|
35
|
+
return {
|
|
36
|
+
path: path.join(globalAgentsDir, 'skills'),
|
|
37
|
+
label: '~/.agents/skills/',
|
|
38
|
+
manifestPath: manifest.getManifestPath(globalAgentsDir),
|
|
39
|
+
};
|
|
40
|
+
}
|
|
41
|
+
return null;
|
|
42
|
+
},
|
|
43
|
+
install({ isGlobal, targetWorkspaceRoot, advisory }) {
|
|
44
|
+
if (!isGlobal) {
|
|
45
|
+
const targetFile = path.join(targetWorkspaceRoot, '.hermes.md');
|
|
46
|
+
advisory.injectAdvisoryText(targetFile, '# .hermes.md', '.hermes.md');
|
|
47
|
+
} else {
|
|
48
|
+
console.log(` ℹ️ Hermes Agent has no documented global instructions file (it reads .hermes.md from the current project directory only) - skipping global install for --hermes.`);
|
|
49
|
+
}
|
|
50
|
+
},
|
|
51
|
+
uninstall({ removeLocal, removeGlobal, workspaceRoot, userHome }) {
|
|
52
|
+
const advisory = require('../../../../scripts/lib/advisory-text');
|
|
53
|
+
if (removeLocal) {
|
|
54
|
+
advisory.removeAdvisoryText(path.join(workspaceRoot, '.hermes.md'));
|
|
55
|
+
}
|
|
56
|
+
}
|
|
57
|
+
};
|
|
@@ -1,17 +1,17 @@
|
|
|
1
|
-
const claude = require('./claude');
|
|
2
|
-
const cursor = require('./cursor');
|
|
3
|
-
const copilot = require('./copilot');
|
|
4
|
-
const codex = require('./codex');
|
|
5
|
-
const _continue = require('./continue');
|
|
6
|
-
const hermes = require('./hermes');
|
|
7
|
-
const worktrees = require('./worktrees');
|
|
8
|
-
|
|
9
|
-
module.exports = [
|
|
10
|
-
claude,
|
|
11
|
-
cursor,
|
|
12
|
-
copilot,
|
|
13
|
-
codex,
|
|
14
|
-
_continue,
|
|
15
|
-
hermes,
|
|
16
|
-
worktrees
|
|
17
|
-
];
|
|
1
|
+
const claude = require('./claude');
|
|
2
|
+
const cursor = require('./cursor');
|
|
3
|
+
const copilot = require('./copilot');
|
|
4
|
+
const codex = require('./codex');
|
|
5
|
+
const _continue = require('./continue');
|
|
6
|
+
const hermes = require('./hermes');
|
|
7
|
+
const worktrees = require('./worktrees');
|
|
8
|
+
|
|
9
|
+
module.exports = [
|
|
10
|
+
claude,
|
|
11
|
+
cursor,
|
|
12
|
+
copilot,
|
|
13
|
+
codex,
|
|
14
|
+
_continue,
|
|
15
|
+
hermes,
|
|
16
|
+
worktrees
|
|
17
|
+
];
|
|
@@ -1,20 +1,20 @@
|
|
|
1
|
-
const path = require('path');
|
|
2
|
-
const fs = require('fs');
|
|
3
|
-
|
|
4
|
-
module.exports = {
|
|
5
|
-
name: 'worktrees',
|
|
6
|
-
label: 'Git Worktrees',
|
|
7
|
-
getIgnorePatterns(workspaceRoot) {
|
|
8
|
-
const patterns = [];
|
|
9
|
-
if (fs.existsSync(path.join(workspaceRoot, '.worktrees'))) {
|
|
10
|
-
patterns.push('.worktrees/');
|
|
11
|
-
}
|
|
12
|
-
if (fs.existsSync(path.join(workspaceRoot, 'worktrees'))) {
|
|
13
|
-
patterns.push('worktrees/');
|
|
14
|
-
}
|
|
15
|
-
return patterns;
|
|
16
|
-
},
|
|
17
|
-
isInstalled() {
|
|
18
|
-
return false; // Not a standalone platform hook/settings installer
|
|
19
|
-
}
|
|
20
|
-
};
|
|
1
|
+
const path = require('path');
|
|
2
|
+
const fs = require('fs');
|
|
3
|
+
|
|
4
|
+
module.exports = {
|
|
5
|
+
name: 'worktrees',
|
|
6
|
+
label: 'Git Worktrees',
|
|
7
|
+
getIgnorePatterns(workspaceRoot) {
|
|
8
|
+
const patterns = [];
|
|
9
|
+
if (fs.existsSync(path.join(workspaceRoot, '.worktrees'))) {
|
|
10
|
+
patterns.push('.worktrees/');
|
|
11
|
+
}
|
|
12
|
+
if (fs.existsSync(path.join(workspaceRoot, 'worktrees'))) {
|
|
13
|
+
patterns.push('worktrees/');
|
|
14
|
+
}
|
|
15
|
+
return patterns;
|
|
16
|
+
},
|
|
17
|
+
isInstalled() {
|
|
18
|
+
return false; // Not a standalone platform hook/settings installer
|
|
19
|
+
}
|
|
20
|
+
};
|
|
@@ -1,30 +1,30 @@
|
|
|
1
|
-
#!/usr/bin/env node
|
|
2
|
-
// Manual escape hatch for the Rule of 3 circuit breaker. Run this directly
|
|
3
|
-
// in your own terminal (not through the agent — Bash is blocked while the
|
|
4
|
-
// breaker is tripped) to clear it without waiting for a new session.
|
|
5
|
-
const fs = require('fs');
|
|
6
|
-
const path = require('path');
|
|
7
|
-
const { getWorkspaceRoot, getSessionDir, readCurrentSession } = require('./lib/harness-state');
|
|
8
|
-
|
|
9
|
-
// This is run directly by the human, not through the agent, so there's no
|
|
10
|
-
// hook payload to read a session_id from - resolve "the session that's
|
|
11
|
-
// currently blocked" via the pointer bootstrap.js wrote at its last
|
|
12
|
-
// SessionStart.
|
|
13
|
-
const root = getWorkspaceRoot();
|
|
14
|
-
const harnessDir = getSessionDir(root, readCurrentSession(root));
|
|
15
|
-
const stateFile = path.join(harnessDir, 'rule-of-3-state.json');
|
|
16
|
-
const reportFile = path.join(harnessDir, 'zoom-out-report.md');
|
|
17
|
-
|
|
18
|
-
if (fs.existsSync(stateFile)) {
|
|
19
|
-
fs.unlinkSync(stateFile);
|
|
20
|
-
console.log('Rule of 3 circuit breaker cleared.');
|
|
21
|
-
} else {
|
|
22
|
-
console.log('Rule of 3 circuit breaker is not currently tripped.');
|
|
23
|
-
}
|
|
24
|
-
|
|
25
|
-
// A human reset is a full clear: the stale reflection report belongs to the
|
|
26
|
-
// loop that was just abandoned.
|
|
27
|
-
if (fs.existsSync(reportFile)) {
|
|
28
|
-
fs.unlinkSync(reportFile);
|
|
29
|
-
console.log('Stale zoom-out report removed.');
|
|
30
|
-
}
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
// Manual escape hatch for the Rule of 3 circuit breaker. Run this directly
|
|
3
|
+
// in your own terminal (not through the agent — Bash is blocked while the
|
|
4
|
+
// breaker is tripped) to clear it without waiting for a new session.
|
|
5
|
+
const fs = require('fs');
|
|
6
|
+
const path = require('path');
|
|
7
|
+
const { getWorkspaceRoot, getSessionDir, readCurrentSession } = require('./lib/harness-state');
|
|
8
|
+
|
|
9
|
+
// This is run directly by the human, not through the agent, so there's no
|
|
10
|
+
// hook payload to read a session_id from - resolve "the session that's
|
|
11
|
+
// currently blocked" via the pointer bootstrap.js wrote at its last
|
|
12
|
+
// SessionStart.
|
|
13
|
+
const root = getWorkspaceRoot();
|
|
14
|
+
const harnessDir = getSessionDir(root, readCurrentSession(root));
|
|
15
|
+
const stateFile = path.join(harnessDir, 'rule-of-3-state.json');
|
|
16
|
+
const reportFile = path.join(harnessDir, 'zoom-out-report.md');
|
|
17
|
+
|
|
18
|
+
if (fs.existsSync(stateFile)) {
|
|
19
|
+
fs.unlinkSync(stateFile);
|
|
20
|
+
console.log('Rule of 3 circuit breaker cleared.');
|
|
21
|
+
} else {
|
|
22
|
+
console.log('Rule of 3 circuit breaker is not currently tripped.');
|
|
23
|
+
}
|
|
24
|
+
|
|
25
|
+
// A human reset is a full clear: the stale reflection report belongs to the
|
|
26
|
+
// loop that was just abandoned.
|
|
27
|
+
if (fs.existsSync(reportFile)) {
|
|
28
|
+
fs.unlinkSync(reportFile);
|
|
29
|
+
console.log('Stale zoom-out report removed.');
|
|
30
|
+
}
|
|
@@ -1,91 +1,124 @@
|
|
|
1
|
-
#!/usr/bin/env node
|
|
2
|
-
const fs = require('fs');
|
|
3
|
-
const path = require('path');
|
|
4
|
-
const crypto = require('crypto');
|
|
5
|
-
const { getWorkspaceRoot, getSessionDir } = require('./lib/harness-state');
|
|
6
|
-
|
|
7
|
-
// PostToolUse hooks in Claude Code typically receive the tool output via stdin
|
|
8
|
-
let inputData = '';
|
|
9
|
-
|
|
10
|
-
process.stdin.on('data', chunk => {
|
|
11
|
-
inputData += chunk;
|
|
12
|
-
});
|
|
13
|
-
|
|
14
|
-
process.stdin.on('end', () => {
|
|
15
|
-
try {
|
|
16
|
-
const payload = JSON.parse(inputData);
|
|
17
|
-
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
|
|
23
|
-
const
|
|
24
|
-
const
|
|
25
|
-
const
|
|
26
|
-
const
|
|
27
|
-
const
|
|
28
|
-
|
|
29
|
-
|
|
30
|
-
const
|
|
31
|
-
|
|
32
|
-
|
|
33
|
-
|
|
34
|
-
|
|
35
|
-
|
|
36
|
-
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
|
|
45
|
-
|
|
46
|
-
|
|
47
|
-
|
|
48
|
-
|
|
49
|
-
|
|
50
|
-
|
|
51
|
-
|
|
52
|
-
|
|
53
|
-
|
|
54
|
-
|
|
55
|
-
|
|
56
|
-
|
|
57
|
-
|
|
58
|
-
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
62
|
-
|
|
63
|
-
|
|
64
|
-
|
|
65
|
-
|
|
66
|
-
|
|
67
|
-
|
|
68
|
-
|
|
69
|
-
|
|
70
|
-
|
|
71
|
-
|
|
72
|
-
|
|
73
|
-
|
|
74
|
-
|
|
75
|
-
|
|
76
|
-
|
|
77
|
-
|
|
78
|
-
|
|
79
|
-
if (
|
|
80
|
-
|
|
81
|
-
|
|
82
|
-
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
|
|
90
|
-
|
|
91
|
-
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
const fs = require('fs');
|
|
3
|
+
const path = require('path');
|
|
4
|
+
const crypto = require('crypto');
|
|
5
|
+
const { getWorkspaceRoot, getSessionDir } = require('./lib/harness-state');
|
|
6
|
+
|
|
7
|
+
// PostToolUse hooks in Claude Code typically receive the tool output via stdin
|
|
8
|
+
let inputData = '';
|
|
9
|
+
|
|
10
|
+
process.stdin.on('data', chunk => {
|
|
11
|
+
inputData += chunk;
|
|
12
|
+
});
|
|
13
|
+
|
|
14
|
+
process.stdin.on('end', () => {
|
|
15
|
+
try {
|
|
16
|
+
const payload = JSON.parse(inputData);
|
|
17
|
+
const toolResponse = payload.tool_response || {};
|
|
18
|
+
const stdout = toolResponse.stdout ?? payload.stdout ?? '';
|
|
19
|
+
const stderr = toolResponse.stderr ?? payload.stderr ?? '';
|
|
20
|
+
const rawExitCode = toolResponse.exitCode ?? toolResponse.exit_code ?? payload.exitCode;
|
|
21
|
+
const exitCode = typeof rawExitCode === 'number' ? rawExitCode : undefined;
|
|
22
|
+
|
|
23
|
+
const explicitFailure = exitCode !== undefined && exitCode !== 0;
|
|
24
|
+
const explicitSuccess = exitCode === 0;
|
|
25
|
+
const stderrSignal = typeof stderr === 'string' && stderr.trim().length > 0;
|
|
26
|
+
const looksLikeError = /\b(error|fail|failed|failure|exception|fatal|panic|traceback|denied|refused|cannot|unable)\b/i.test(stderr);
|
|
27
|
+
const isFailure = explicitFailure || (exitCode === undefined && stderrSignal && looksLikeError);
|
|
28
|
+
const errorText = (stderrSignal ? stderr : stdout) || '';
|
|
29
|
+
|
|
30
|
+
const stateFile = path.join(getSessionDir(getWorkspaceRoot(), payload.session_id), 'rule-of-3-state.json');
|
|
31
|
+
|
|
32
|
+
let state = { count: 0, lastHash: null, zoomOutResolved: false, zoomOutCycles: 0, lastFailureAt: 0 };
|
|
33
|
+
if (fs.existsSync(stateFile)) {
|
|
34
|
+
state = JSON.parse(fs.readFileSync(stateFile, 'utf8'));
|
|
35
|
+
}
|
|
36
|
+
|
|
37
|
+
if (isFailure && errorText) {
|
|
38
|
+
// Enhanced failure signature detection with context-aware categorization
|
|
39
|
+
// Hash includes: error message, file path, command attempted, and failure category
|
|
40
|
+
const filePath = payload.tool_input?.file_path || payload.tool_input?.filePath || '';
|
|
41
|
+
const command = payload.tool_input?.command || '';
|
|
42
|
+
|
|
43
|
+
// Categorize failure type for context-aware thresholds
|
|
44
|
+
let category = 'unknown';
|
|
45
|
+
if (/SyntaxError|Unexpected token|Syntax Error/i.test(errorText)) {
|
|
46
|
+
category = 'syntax';
|
|
47
|
+
} else if (/ENOENT|not found|No such file|找不到|不存在/i.test(errorText)) {
|
|
48
|
+
category = 'environment';
|
|
49
|
+
} else if (/timeout|timed out|超時|逾時/i.test(errorText)) {
|
|
50
|
+
category = 'timeout';
|
|
51
|
+
} else if (/test.*fail|assertion|AssertionError|測試.*失敗|斷言.*錯誤/i.test(errorText)) {
|
|
52
|
+
category = 'test';
|
|
53
|
+
} else if (/permission|denied|EACCES|權限/i.test(errorText)) {
|
|
54
|
+
category = 'permission';
|
|
55
|
+
} else if (/dependency|module not found|Cannot find module|模組.*找不到/i.test(errorText)) {
|
|
56
|
+
category = 'dependency';
|
|
57
|
+
}
|
|
58
|
+
|
|
59
|
+
// Normalize error text with context for better loop detection
|
|
60
|
+
const normalized = errorText.slice(-400)
|
|
61
|
+
.toLowerCase()
|
|
62
|
+
.replace(/0x[0-9a-f]+/g, '#')
|
|
63
|
+
.replace(/\d+/g, '#')
|
|
64
|
+
.replace(/\s+/g, ' ')
|
|
65
|
+
.trim();
|
|
66
|
+
|
|
67
|
+
// Include file path and command in hash for better context
|
|
68
|
+
const contextString = `${category}:${filePath}:${command}:${normalized.slice(-200)}`;
|
|
69
|
+
const hash = crypto.createHash('md5').update(contextString).digest('hex');
|
|
70
|
+
|
|
71
|
+
// Context-aware thresholds: different failure types have different breaking points
|
|
72
|
+
let threshold = 3; // Default: 3 strikes
|
|
73
|
+
if (category === 'syntax') {
|
|
74
|
+
threshold = 3; // Same syntax error 3 times = stop
|
|
75
|
+
} else if (category === 'environment') {
|
|
76
|
+
threshold = 4; // Environment issues may need more exploration
|
|
77
|
+
} else if (category === 'test') {
|
|
78
|
+
threshold = 3; // Same test failure 3 times = stop
|
|
79
|
+
} else if (category === 'timeout') {
|
|
80
|
+
threshold = 2; // Timeouts are expensive, fail faster
|
|
81
|
+
} else if (category === 'permission') {
|
|
82
|
+
threshold = 2; // Permission issues rarely resolve by retrying
|
|
83
|
+
} else if (category === 'dependency') {
|
|
84
|
+
threshold = 4; // Dependency issues may need different approaches
|
|
85
|
+
}
|
|
86
|
+
|
|
87
|
+
if (state.lastHash === hash) {
|
|
88
|
+
state.count += 1;
|
|
89
|
+
state.category = category; // Track failure category for smarter decisions
|
|
90
|
+
} else {
|
|
91
|
+
state.count = 1;
|
|
92
|
+
state.lastHash = hash;
|
|
93
|
+
state.category = category;
|
|
94
|
+
// New signature = new problem: the zoom-out cycle budget starts over.
|
|
95
|
+
state.zoomOutCycles = 0;
|
|
96
|
+
}
|
|
97
|
+
|
|
98
|
+
// Store threshold for this failure type
|
|
99
|
+
state.threshold = threshold;
|
|
100
|
+
|
|
101
|
+
// Any fresh failure re-arms the breaker. Without this, a signature that
|
|
102
|
+
// failed again after a success (or after a zoom-out release) would keep
|
|
103
|
+
// zoomOutResolved=true and the breaker could never trip on it again.
|
|
104
|
+
state.zoomOutResolved = false;
|
|
105
|
+
// rule-of-3.js only honors a zoom-out report written AFTER this moment,
|
|
106
|
+
// so a stale report can't unlock a later trip.
|
|
107
|
+
state.lastFailureAt = Date.now();
|
|
108
|
+
|
|
109
|
+
fs.writeFileSync(stateFile, JSON.stringify(state, null, 2), 'utf8');
|
|
110
|
+
} else if (explicitSuccess) {
|
|
111
|
+
// Only reset on a *confirmed* zero exit code, not merely "not a failure".
|
|
112
|
+
if (state.count > 0 || state.zoomOutCycles > 0) {
|
|
113
|
+
state.count = 0;
|
|
114
|
+
state.zoomOutResolved = true;
|
|
115
|
+
state.zoomOutCycles = 0;
|
|
116
|
+
fs.writeFileSync(stateFile, JSON.stringify(state, null, 2), 'utf8');
|
|
117
|
+
}
|
|
118
|
+
}
|
|
119
|
+
|
|
120
|
+
} catch (err) {
|
|
121
|
+
// silently fail
|
|
122
|
+
}
|
|
123
|
+
process.exit(0);
|
|
124
|
+
});
|