harness-everything 0.2.1 → 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 +63 -63
- package/docs/workflows/eval-harness.md +64 -57
- package/docs/workflows/fable-discipline.md +59 -59
- package/docs/workflows/fable-mode.md +12 -15
- package/docs/workflows/find-skills.md +36 -0
- package/docs/workflows/git-commit.md +71 -62
- package/docs/workflows/grill-me.md +62 -62
- package/docs/workflows/grill-with-docs.md +69 -62
- package/docs/workflows/harness-everything.md +95 -92
- package/docs/workflows/improve-codebase-architecture.md +68 -60
- package/docs/workflows/install-cognitive-os.md +60 -60
- package/docs/workflows/multi-agent-workspace.md +42 -0
- package/docs/workflows/repo-docs.md +64 -59
- package/docs/workflows/rewrite-commits.md +60 -60
- package/docs/workflows/security-review.md +77 -58
- package/docs/workflows/self-evolve.md +42 -59
- package/docs/workflows/skill-creator.md +80 -75
- package/docs/workflows/skill-style.md +64 -64
- package/docs/workflows/tdd.md +7 -4
- package/docs/workflows/to-spec.md +83 -0
- package/docs/workflows/to-tickets.md +80 -0
- package/docs/workflows/todo-driven-workflow.md +28 -30
- package/docs/workflows/using-git-worktrees.md +67 -63
- package/docs/workflows/verification-loop.md +71 -69
- package/docs/workflows/verify-before-claim.md +64 -60
- package/docs/workflows/zoom-out.md +74 -62
- package/environment-detection/SKILL.md +41 -77
- 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 +42 -59
- 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 +41 -38
- package/fable-discipline/references/discipline-rules.md +22 -0
- package/fable-mode/CONTRACT-FORMAT.md +51 -51
- package/fable-mode/SKILL.md +39 -51
- 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 +45 -55
- 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 +43 -36
- 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 +41 -99
- package/grill-with-docs/references/session-playbook.md +97 -0
- package/harness-everything/SKILL.md +42 -141
- 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 +87 -18
- package/hooks/hooks.json +155 -155
- package/hooks/scripts/boundary-guard.js +55 -55
- 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 +185 -42
- package/hooks/scripts/lib/platforms/codex.js +114 -46
- package/hooks/scripts/lib/platforms/continue.js +91 -41
- package/hooks/scripts/lib/platforms/copilot.js +109 -41
- package/hooks/scripts/lib/platforms/cursor.js +89 -41
- package/hooks/scripts/lib/platforms/hermes.js +57 -32
- 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 +14 -3
- 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 +40 -44
- 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 +38 -95
- 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 +46 -45
- 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 +37 -32
- package/rewrite-commits/references/workflow.md +18 -0
- package/scripts/installer.js +546 -649
- package/scripts/lib/advisory-text.js +167 -163
- package/scripts/lib/claude-hooks.js +84 -71
- package/scripts/lib/gitignore.js +199 -80
- package/scripts/lib/manifest.js +146 -82
- package/scripts/lib/prompts.js +241 -241
- package/scripts/lib/skills.js +260 -172
- package/scripts/lib/workspace.js +28 -28
- package/scripts/verify-install.js +147 -0
- package/security-review/SKILL.md +50 -511
- 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 +47 -54
- 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 +49 -91
- 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 +20 -25
- package/skill-style/SKILL.md +42 -48
- package/skill-style/references/style-guide.md +36 -0
- package/tdd/SKILL.md +44 -53
- 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 +40 -54
- package/todo-driven-workflow/references/execution-guide.md +27 -0
- package/using-git-worktrees/SKILL.md +49 -161
- package/using-git-worktrees/references/workflow-details.md +108 -0
- package/verification-loop/SKILL.md +46 -138
- 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 +43 -86
- package/verify-before-claim/references/verification-guide.md +73 -0
- package/zoom-out/SKILL.md +42 -90
- 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
|
@@ -0,0 +1,316 @@
|
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
/**
|
|
3
|
+
* Behavioral Eval Runner with opencode plugin
|
|
4
|
+
*
|
|
5
|
+
* Runs discipline cases with opencode plugin hooks enabled.
|
|
6
|
+
* Usage: node behavioral-evals/run-with-plugin.js run [--case <id>]
|
|
7
|
+
*/
|
|
8
|
+
|
|
9
|
+
const fs = require('fs');
|
|
10
|
+
const os = require('os');
|
|
11
|
+
const path = require('path');
|
|
12
|
+
const { execFileSync } = require('child_process');
|
|
13
|
+
|
|
14
|
+
const ROOT = path.resolve(__dirname, '..');
|
|
15
|
+
const CASES_DIR = path.join(__dirname, 'cases');
|
|
16
|
+
const RESULTS_DIR = path.join(__dirname, 'results');
|
|
17
|
+
const PLUGIN_DIR = path.join(ROOT, 'opencode-plugin');
|
|
18
|
+
|
|
19
|
+
// Minimal YAML subset parser (copied from run.js)
|
|
20
|
+
function parseSimpleYaml(text) {
|
|
21
|
+
const lines = text.split('\n');
|
|
22
|
+
let i = 0;
|
|
23
|
+
function parseBlock(indent) {
|
|
24
|
+
const obj = {};
|
|
25
|
+
const arr = [];
|
|
26
|
+
let isArray = null;
|
|
27
|
+
while (i < lines.length) {
|
|
28
|
+
const line = lines[i];
|
|
29
|
+
if (!line.trim() || line.trim().startsWith('#')) { i++; continue; }
|
|
30
|
+
const currentIndent = line.match(/^ */)[0].length;
|
|
31
|
+
if (currentIndent < indent) break;
|
|
32
|
+
if (currentIndent > indent) { i++; continue; }
|
|
33
|
+
const listItem = line.trim().startsWith('- ');
|
|
34
|
+
if (isArray === null) isArray = listItem;
|
|
35
|
+
else if (isArray !== listItem) throw new Error('Mixed list/map at same indent near: ' + line);
|
|
36
|
+
if (listItem) {
|
|
37
|
+
const content = line.trim().slice(2);
|
|
38
|
+
const kv = content.match(/^([^:]+):\s*(.*)$/);
|
|
39
|
+
if (kv && !content.startsWith('"') && !content.startsWith("'")) {
|
|
40
|
+
const item = { [kv[1].trim()]: scalar(kv[2]) };
|
|
41
|
+
i++;
|
|
42
|
+
Object.assign(item, parseNested(currentIndent + 2));
|
|
43
|
+
arr.push(item);
|
|
44
|
+
} else {
|
|
45
|
+
arr.push(scalar(content));
|
|
46
|
+
i++;
|
|
47
|
+
}
|
|
48
|
+
} else {
|
|
49
|
+
const kv = line.match(/^([^:#]+):\s*(.*)$/);
|
|
50
|
+
if (!kv) { i++; continue; }
|
|
51
|
+
const key = kv[1].trim();
|
|
52
|
+
const rawVal = kv[2];
|
|
53
|
+
if (rawVal === '|' || rawVal === '>') {
|
|
54
|
+
i++;
|
|
55
|
+
let block = [];
|
|
56
|
+
while (i < lines.length) {
|
|
57
|
+
const l = lines[i];
|
|
58
|
+
const ind = l.match(/^ */)[0].length;
|
|
59
|
+
if (!l.trim()) { block.push(''); i++; continue; }
|
|
60
|
+
if (ind <= indent) break;
|
|
61
|
+
block.push(l.slice(indent + 2));
|
|
62
|
+
i++;
|
|
63
|
+
}
|
|
64
|
+
obj[key] = block.join('\n');
|
|
65
|
+
} else if (rawVal === '') {
|
|
66
|
+
i++;
|
|
67
|
+
obj[key] = parseNested(indent + 2);
|
|
68
|
+
} else {
|
|
69
|
+
obj[key] = scalar(rawVal);
|
|
70
|
+
i++;
|
|
71
|
+
}
|
|
72
|
+
}
|
|
73
|
+
}
|
|
74
|
+
return isArray ? arr : obj;
|
|
75
|
+
}
|
|
76
|
+
function parseNested(minIndent) {
|
|
77
|
+
let j = i;
|
|
78
|
+
while (j < lines.length && (!lines[j].trim() || lines[j].trim().startsWith('#'))) j++;
|
|
79
|
+
if (j >= lines.length) return null;
|
|
80
|
+
const ind = lines[j].match(/^ */)[0].length;
|
|
81
|
+
const block = parseBlock(ind);
|
|
82
|
+
return block;
|
|
83
|
+
}
|
|
84
|
+
function scalar(v) {
|
|
85
|
+
const t = v.trim();
|
|
86
|
+
if (t === 'true') return true;
|
|
87
|
+
if (t === 'false') return false;
|
|
88
|
+
if (/^-?\d+$/.test(t)) return parseInt(t, 10);
|
|
89
|
+
if ((t.startsWith('"') && t.endsWith('"')) || (t.startsWith("'") && t.endsWith("'"))) return t.slice(1, -1);
|
|
90
|
+
return t;
|
|
91
|
+
}
|
|
92
|
+
const result = parseBlock(0);
|
|
93
|
+
return result || {};
|
|
94
|
+
}
|
|
95
|
+
|
|
96
|
+
function discoverCases() {
|
|
97
|
+
return fs.readdirSync(CASES_DIR)
|
|
98
|
+
.filter((f) => f.endsWith('.yaml'))
|
|
99
|
+
.map((f) => ({ file: path.join(CASES_DIR, f), ...parseSimpleYaml(fs.readFileSync(path.join(CASES_DIR, f), 'utf8')) }));
|
|
100
|
+
}
|
|
101
|
+
|
|
102
|
+
function fail(msg) {
|
|
103
|
+
console.error(`❌ ${msg}`);
|
|
104
|
+
process.exit(1);
|
|
105
|
+
}
|
|
106
|
+
|
|
107
|
+
function buildWorkspace(c) {
|
|
108
|
+
const ws = fs.mkdtempSync(path.join(os.tmpdir(), `harness-behavioral-plugin-${c.id}-`));
|
|
109
|
+
for (const f of c.fixture.files) {
|
|
110
|
+
const target = path.join(ws, f.path);
|
|
111
|
+
fs.mkdirSync(path.dirname(target), { recursive: true });
|
|
112
|
+
fs.writeFileSync(target, typeof f.content === 'string' ? f.content.replace(/\n$/, '') + '\n' : String(f.content));
|
|
113
|
+
}
|
|
114
|
+
return ws;
|
|
115
|
+
}
|
|
116
|
+
|
|
117
|
+
function gitSnapshot(ws) {
|
|
118
|
+
execFileSync('git', ['init', '-q'], { cwd: ws });
|
|
119
|
+
execFileSync('git', ['config', 'user.email', 'eval@harness.local'], { cwd: ws });
|
|
120
|
+
execFileSync('git', ['config', 'user.name', 'Harness Eval'], { cwd: ws });
|
|
121
|
+
execFileSync('git', ['add', '.'], { cwd: ws });
|
|
122
|
+
execFileSync('git', ['commit', '-qm', 'fixture'], { cwd: ws });
|
|
123
|
+
}
|
|
124
|
+
|
|
125
|
+
function installHarnessWithPlugin(ws) {
|
|
126
|
+
const src = path.join(ws, '.harness-src');
|
|
127
|
+
fs.mkdirSync(src, { recursive: true });
|
|
128
|
+
for (const entry of fs.readdirSync(ROOT, { withFileTypes: true })) {
|
|
129
|
+
if (entry.name.startsWith('.') || entry.name === 'node_modules' || entry.name === 'benchmarks' || entry.name === 'behavioral-evals') continue;
|
|
130
|
+
if (!entry.isDirectory()) continue;
|
|
131
|
+
fs.cpSync(path.join(ROOT, entry.name), path.join(src, entry.name), { recursive: true });
|
|
132
|
+
}
|
|
133
|
+
|
|
134
|
+
// Skills where Claude Code can see them
|
|
135
|
+
const skillsDir = path.join(ws, '.claude', 'skills');
|
|
136
|
+
fs.mkdirSync(skillsDir, { recursive: true });
|
|
137
|
+
const skillNames = [];
|
|
138
|
+
for (const entry of fs.readdirSync(src, { withFileTypes: true })) {
|
|
139
|
+
if (!entry.isDirectory()) continue;
|
|
140
|
+
if (!fs.existsSync(path.join(src, entry.name, 'SKILL.md'))) continue;
|
|
141
|
+
fs.cpSync(path.join(src, entry.name), path.join(skillsDir, entry.name), { recursive: true });
|
|
142
|
+
skillNames.push(entry.name);
|
|
143
|
+
}
|
|
144
|
+
|
|
145
|
+
// Hooks: rewrite relative commands to the copied source tree
|
|
146
|
+
const hooks = JSON.parse(fs.readFileSync(path.join(src, 'hooks', 'hooks.json'), 'utf8'));
|
|
147
|
+
const hooksCfg = hooks.hooks || {};
|
|
148
|
+
for (const events of Object.values(hooksCfg)) {
|
|
149
|
+
for (const entry of events) {
|
|
150
|
+
for (const h of entry.hooks || []) {
|
|
151
|
+
h.command = h.command.replace(/^node /, `node "${src}/`) + '"';
|
|
152
|
+
}
|
|
153
|
+
}
|
|
154
|
+
}
|
|
155
|
+
const settingsPath = path.join(ws, '.claude', 'settings.json');
|
|
156
|
+
const settings = fs.existsSync(settingsPath) ? JSON.parse(fs.readFileSync(settingsPath, 'utf8')) : {};
|
|
157
|
+
settings.hooks = { ...(settings.hooks || {}), ...hooksCfg };
|
|
158
|
+
fs.mkdirSync(path.dirname(settingsPath), { recursive: true });
|
|
159
|
+
fs.writeFileSync(settingsPath, JSON.stringify(settings, null, 2));
|
|
160
|
+
|
|
161
|
+
// opencode config with skills AND plugin hooks
|
|
162
|
+
const pluginHooks = {};
|
|
163
|
+
const pluginJson = JSON.parse(fs.readFileSync(path.join(PLUGIN_DIR, 'plugin.json'), 'utf8'));
|
|
164
|
+
|
|
165
|
+
// Map plugin hooks to opencode hook events
|
|
166
|
+
if (pluginJson.hooks) {
|
|
167
|
+
for (const [event, hookFile] of Object.entries(pluginJson.hooks)) {
|
|
168
|
+
const hookPath = path.join(PLUGIN_DIR, hookFile);
|
|
169
|
+
if (fs.existsSync(hookPath)) {
|
|
170
|
+
pluginHooks[event] = [{ type: 'command', command: `node "${hookPath}"` }];
|
|
171
|
+
}
|
|
172
|
+
}
|
|
173
|
+
}
|
|
174
|
+
|
|
175
|
+
fs.writeFileSync(
|
|
176
|
+
path.join(ws, 'opencode.json'),
|
|
177
|
+
JSON.stringify({
|
|
178
|
+
$schema: 'https://opencode.ai/config.json',
|
|
179
|
+
instructions: skillNames.map((n) => `.claude/skills/${n}/SKILL.md`),
|
|
180
|
+
hooks: pluginHooks
|
|
181
|
+
}, null, 2)
|
|
182
|
+
);
|
|
183
|
+
|
|
184
|
+
// Also copy plugin hooks to workspace for reference
|
|
185
|
+
fs.cpSync(PLUGIN_DIR, path.join(src, 'opencode-plugin'), { recursive: true });
|
|
186
|
+
}
|
|
187
|
+
|
|
188
|
+
function runHeadless(prompt, ws, maxTurns) {
|
|
189
|
+
const outPath = path.join(path.dirname(ws), path.basename(ws) + '.transcript.json');
|
|
190
|
+
const model = process.env.BEHAVIORAL_MODEL;
|
|
191
|
+
const args = ['run', '--format', 'json', '--auto', '--dir', ws];
|
|
192
|
+
if (model) args.push('-m', model);
|
|
193
|
+
// Pass the prompt as one argv element; shell quoting corrupts prompts on
|
|
194
|
+
// Windows and previously reduced full requests to fragments such as `Add.`.
|
|
195
|
+
args.push(prompt);
|
|
196
|
+
const outputFd = fs.openSync(outPath, 'w');
|
|
197
|
+
try {
|
|
198
|
+
execFileSync('opencode', args,
|
|
199
|
+
{ cwd: ws, stdio: ['ignore', outputFd, 'inherit'], timeout: 20 * 60 * 1000, windowsHide: true });
|
|
200
|
+
} finally {
|
|
201
|
+
fs.closeSync(outputFd);
|
|
202
|
+
}
|
|
203
|
+
return outPath;
|
|
204
|
+
}
|
|
205
|
+
|
|
206
|
+
function extractAgentTrace(transcriptPath) {
|
|
207
|
+
const raw = fs.readFileSync(transcriptPath, 'utf8');
|
|
208
|
+
const chunks = [];
|
|
209
|
+
let structured = false;
|
|
210
|
+
for (const line of raw.split('\n')) {
|
|
211
|
+
if (!line.trim()) continue;
|
|
212
|
+
let e;
|
|
213
|
+
try { e = JSON.parse(line); } catch { continue; }
|
|
214
|
+
const p = e && e.part;
|
|
215
|
+
if (!p) continue;
|
|
216
|
+
if (e.type === 'text' && typeof p.text === 'string') {
|
|
217
|
+
chunks.push(p.text);
|
|
218
|
+
structured = true;
|
|
219
|
+
} else if (String(e.type).includes('tool')) {
|
|
220
|
+
const input = p.state && p.state.input !== undefined ? p.state.input : p.input;
|
|
221
|
+
chunks.push(`[${p.tool || 'tool'}] ` + JSON.stringify(input ?? {}));
|
|
222
|
+
structured = true;
|
|
223
|
+
}
|
|
224
|
+
}
|
|
225
|
+
if (!structured) {
|
|
226
|
+
try {
|
|
227
|
+
const obj = JSON.parse(raw);
|
|
228
|
+
if (obj && typeof obj.result === 'string') return obj.result;
|
|
229
|
+
} catch { /* fall through */ }
|
|
230
|
+
return raw;
|
|
231
|
+
}
|
|
232
|
+
return chunks.join('\n');
|
|
233
|
+
}
|
|
234
|
+
|
|
235
|
+
function grade(c, ws, transcriptPath) {
|
|
236
|
+
const trace = extractAgentTrace(transcriptPath);
|
|
237
|
+
const results = [];
|
|
238
|
+
for (const e of c.expectations) {
|
|
239
|
+
let pass = false;
|
|
240
|
+
try {
|
|
241
|
+
if (e.type === 'trace_contains') pass = trace.includes(e.value);
|
|
242
|
+
else if (e.type === 'trace_not_contains') pass = !trace.includes(e.value);
|
|
243
|
+
else if (e.type === 'file_contains') pass = fs.readFileSync(path.join(ws, e.path), 'utf8').includes(e.value);
|
|
244
|
+
else if (e.type === 'file_not_exists') pass = !fs.existsSync(path.join(ws, e.path));
|
|
245
|
+
else if (e.type === 'command_exit_0') {
|
|
246
|
+
if (process.platform === 'win32') {
|
|
247
|
+
execFileSync(process.env.ComSpec || 'cmd.exe', ['/d', '/s', '/c', e.command], { cwd: ws, stdio: 'ignore' });
|
|
248
|
+
} else {
|
|
249
|
+
execFileSync('/bin/sh', ['-c', e.command], { cwd: ws, stdio: 'ignore' });
|
|
250
|
+
}
|
|
251
|
+
pass = true;
|
|
252
|
+
}
|
|
253
|
+
} catch (err) {
|
|
254
|
+
pass = false;
|
|
255
|
+
}
|
|
256
|
+
results.push({ ...e, pass });
|
|
257
|
+
}
|
|
258
|
+
const gating = results.filter((r) => !r.informational);
|
|
259
|
+
return { results, passed: gating.length > 0 && gating.every((g) => g.pass) };
|
|
260
|
+
}
|
|
261
|
+
|
|
262
|
+
function runLive(filter) {
|
|
263
|
+
console.log('Engine: opencode (with plugin hooks)');
|
|
264
|
+
const cases = discoverCases().filter((c) => !filter || c.id === filter);
|
|
265
|
+
if (filter && cases.length === 0) fail(`no case with id "${filter}"`);
|
|
266
|
+
fs.mkdirSync(RESULTS_DIR, { recursive: true });
|
|
267
|
+
|
|
268
|
+
let failed = 0;
|
|
269
|
+
for (const c of cases) {
|
|
270
|
+
console.log(`\n=== ${c.id}${c.pressure ? ' [PRESSURE]' : ''} ===`);
|
|
271
|
+
const ws = buildWorkspace(c);
|
|
272
|
+
installHarnessWithPlugin(ws);
|
|
273
|
+
if (c.fixture.git) gitSnapshot(ws);
|
|
274
|
+
let transcriptPath;
|
|
275
|
+
try {
|
|
276
|
+
transcriptPath = runHeadless(c.prompt, ws, c.max_turns);
|
|
277
|
+
} catch (err) {
|
|
278
|
+
console.error(`❌ session failed: ${err.message.slice(0, 300)}`);
|
|
279
|
+
failed++;
|
|
280
|
+
continue;
|
|
281
|
+
}
|
|
282
|
+
const { results: graded, passed } = grade(c, ws, transcriptPath);
|
|
283
|
+
const record = {
|
|
284
|
+
id: c.id,
|
|
285
|
+
engine: 'opencode-with-plugin',
|
|
286
|
+
model: process.env.BEHAVIORAL_MODEL || 'openai/gpt-5-mini',
|
|
287
|
+
pressure: !!c.pressure,
|
|
288
|
+
date: new Date().toISOString(),
|
|
289
|
+
workspace: ws,
|
|
290
|
+
transcript: transcriptPath,
|
|
291
|
+
expectations: graded.map(({ description, pass }) => ({ description, pass })),
|
|
292
|
+
outcome: passed ? 'pass' : 'fail',
|
|
293
|
+
};
|
|
294
|
+
const outFile = path.join(RESULTS_DIR, `plugin-${record.date.slice(0, 10)}-${c.id}.json`);
|
|
295
|
+
fs.writeFileSync(outFile, JSON.stringify(record, null, 2));
|
|
296
|
+
for (const g of graded) console.log(` ${g.pass ? '✅' : '❌'}${g.informational ? ' (info)' : ''} ${g.description}`);
|
|
297
|
+
console.log(`${passed ? '✅ PASS' : '❌ FAIL'} -> ${outFile}`);
|
|
298
|
+
if (!passed) failed++;
|
|
299
|
+
}
|
|
300
|
+
console.log(failed ? `\n❌ ${failed} case(s) failed.` : '\n🎉 all cases passed.');
|
|
301
|
+
process.exit(failed ? 1 : 0);
|
|
302
|
+
}
|
|
303
|
+
|
|
304
|
+
const args = process.argv.slice(2);
|
|
305
|
+
function flag(name) {
|
|
306
|
+
const i = args.indexOf(name);
|
|
307
|
+
return i >= 0 ? args[i + 1] : undefined;
|
|
308
|
+
}
|
|
309
|
+
if (args[0] === 'validate') validate(discoverCases());
|
|
310
|
+
else if (args[0] === 'run') {
|
|
311
|
+
const filter = args.includes('--case') ? flag('--case') : undefined;
|
|
312
|
+
runLive(filter);
|
|
313
|
+
} else {
|
|
314
|
+
console.log('Usage:\n node behavioral-evals/run-with-plugin.js run [--case <id>]');
|
|
315
|
+
process.exit(args.length ? 1 : 0);
|
|
316
|
+
}
|