harness-everything 0.3.6 → 0.4.0
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/.agents/plugins/marketplace.json +20 -0
- package/.claude-plugin/marketplace.json +1 -1
- package/.claude-plugin/plugin.json +8 -7
- package/.github/workflows/ci.yml +23 -18
- package/.github/workflows/release.yml +53 -27
- package/.releaserc.json +113 -0
- package/AGENTS.md +34 -23
- package/BENCHMARK_SOP.md +38 -18
- package/CHANGELOG.md +194 -3
- package/CONTRIBUTING.md +17 -4
- package/PRIVACY.md +25 -0
- package/README.md +118 -87
- package/RELEASING.md +82 -0
- package/SUPPORT.md +27 -0
- package/TERMS.md +21 -0
- package/VERIFICATION.md +175 -286
- package/behavioral-evals/README.md +21 -0
- package/behavioral-evals/case-validator.js +194 -0
- package/behavioral-evals/cases/baseline-environment-detection.yaml +21 -0
- package/behavioral-evals/cases/baseline-eval-harness.yaml +27 -0
- package/behavioral-evals/cases/baseline-fable-discipline.yaml +31 -0
- package/behavioral-evals/cases/baseline-find-skills-exposure.yaml +49 -0
- package/behavioral-evals/cases/baseline-find-skills.yaml +21 -0
- package/behavioral-evals/cases/baseline-git-commit.yaml +2 -2
- package/behavioral-evals/cases/baseline-install-cognitive-os.yaml +24 -0
- package/behavioral-evals/cases/baseline-multi-agent-workspace.yaml +28 -0
- package/behavioral-evals/cases/baseline-rewrite-commits.yaml +34 -0
- package/behavioral-evals/cases/baseline-self-evolve.yaml +31 -0
- package/behavioral-evals/cases/baseline-skill-creator.yaml +20 -0
- package/behavioral-evals/cases/baseline-skill-style.yaml +21 -0
- package/behavioral-evals/cases/baseline-to-spec.yaml +34 -0
- package/behavioral-evals/cases/baseline-to-tickets.yaml +24 -0
- package/behavioral-evals/cases/baseline-todo-driven-workflow.yaml +32 -0
- package/behavioral-evals/cases/baseline-using-git-worktrees.yaml +15 -0
- package/behavioral-evals/cases/breaker-zoom-out-after-3.yaml +2 -2
- package/behavioral-evals/cases/pressure-complexity-skip.yaml +2 -2
- package/behavioral-evals/cases/pressure-scope-bypass.yaml +7 -2
- package/behavioral-evals/cases/pressure-sunk-cost-retry.yaml +7 -4
- package/behavioral-evals/cases/scope-discipline.yaml +2 -1
- package/behavioral-evals/cases/verify-before-done.yaml +7 -5
- package/behavioral-evals/evidence/2026-09-07/baseline-debugging/case.yaml +31 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-debugging/historical-case.yaml +31 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-debugging/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-debugging/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-performance/case.yaml +34 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-performance/historical-case.yaml +34 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-performance/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-performance/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-security-review/case.yaml +28 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-security-review/historical-case.yaml +28 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-security-review/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-security-review/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-simple-bugfix/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-simple-bugfix/historical-case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-simple-bugfix/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/baseline-simple-bugfix/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/grill-me-adversarial/case.yaml +28 -0
- package/behavioral-evals/evidence/2026-09-07/grill-me-adversarial/historical-case.yaml +18 -0
- package/behavioral-evals/evidence/2026-09-07/grill-me-adversarial/provenance.json +30 -0
- package/behavioral-evals/evidence/2026-09-07/live-run-blocker.md +16 -0
- package/behavioral-evals/evidence/2026-09-07/manifest.json +88 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-scope-bypass/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-scope-bypass/historical-case.yaml +28 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-scope-bypass/provenance.json +30 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-docs/case.yaml +50 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-docs/historical-case.yaml +50 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-docs/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-docs/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-error-handling/case.yaml +35 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-error-handling/historical-case.yaml +35 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-error-handling/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-error-handling/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification/case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification/historical-case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification/result.sanitized.json +24 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-23-pressure-skip-verification/case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-23-pressure-skip-verification/historical-case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-23-pressure-skip-verification/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-23-pressure-skip-verification/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-27-pressure-skip-verification/case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-27-pressure-skip-verification/historical-case.yaml +43 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-27-pressure-skip-verification/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-skip-verification--2026-08-27-pressure-skip-verification/result.sanitized.json +24 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry/case.yaml +50 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry/historical-case.yaml +47 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry/result.sanitized.json +24 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-23-pressure-sunk-cost-retry/case.yaml +50 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-23-pressure-sunk-cost-retry/historical-case.yaml +47 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-23-pressure-sunk-cost-retry/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-23-pressure-sunk-cost-retry/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-27-pressure-sunk-cost-retry/case.yaml +50 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-27-pressure-sunk-cost-retry/historical-case.yaml +47 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-27-pressure-sunk-cost-retry/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/pressure-sunk-cost-retry--2026-08-27-pressure-sunk-cost-retry/result.sanitized.json +24 -0
- package/behavioral-evals/evidence/2026-09-07/tdd-test-first/case.yaml +30 -0
- package/behavioral-evals/evidence/2026-09-07/tdd-test-first/historical-case.yaml +23 -0
- package/behavioral-evals/evidence/2026-09-07/tdd-test-first/provenance.json +30 -0
- package/behavioral-evals/evidence/2026-09-07/triage-matrix.json +426 -0
- package/behavioral-evals/evidence/2026-09-07/triage-matrix.md +24 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-claim-cites/case.yaml +19 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-claim-cites/historical-case.yaml +18 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-claim-cites/provenance.json +30 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done/historical-case.yaml +31 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-23-verify-before-done/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-23-verify-before-done/historical-case.yaml +31 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-23-verify-before-done/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-23-verify-before-done/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-27-verify-before-done/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-27-verify-before-done/historical-case.yaml +31 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-27-verify-before-done/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-07/verify-before-done--2026-08-27-verify-before-done/result.sanitized.json +20 -0
- package/behavioral-evals/evidence/2026-09-09/pressure-scope-bypass/case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-09/pressure-scope-bypass/historical-case.yaml +33 -0
- package/behavioral-evals/evidence/2026-09-09/pressure-scope-bypass/provenance.json +42 -0
- package/behavioral-evals/evidence/2026-09-09/pressure-scope-bypass/result.sanitized.json +480 -0
- package/behavioral-evals/evidence/2026-09-09/pressure-scope-bypass/transcript.sanitized.jsonl +95 -0
- package/behavioral-evals/evidence-tool.js +443 -0
- package/behavioral-evals/fixtures/claude-denied.jsonl +3 -0
- package/behavioral-evals/fixtures/claude-legacy.json +1 -0
- package/behavioral-evals/fixtures/claude-oral-only.jsonl +2 -0
- package/behavioral-evals/fixtures/claude-success.jsonl +5 -0
- package/behavioral-evals/fixtures/claude-truncated.jsonl +3 -0
- package/behavioral-evals/fixtures/opencode-success.jsonl +6 -0
- package/behavioral-evals/results/2026-08-31-baseline-add-feature.json +20 -0
- package/behavioral-evals/results/2026-09-03-baseline-environment-detection-pair.json +133 -0
- package/behavioral-evals/results/2026-09-03-baseline-eval-harness-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-fable-discipline-pair.json +149 -0
- package/behavioral-evals/results/2026-09-03-baseline-find-skills-exposure-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-find-skills-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-install-cognitive-os-pair.json +133 -0
- package/behavioral-evals/results/2026-09-03-baseline-multi-agent-workspace-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-rewrite-commits-pair.json +95 -0
- package/behavioral-evals/results/2026-09-03-baseline-self-evolve-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-skill-creator-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-skill-style-pair.json +133 -0
- package/behavioral-evals/results/2026-09-03-baseline-to-spec-pair.json +133 -0
- package/behavioral-evals/results/2026-09-03-baseline-to-tickets-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-todo-driven-workflow-pair.json +141 -0
- package/behavioral-evals/results/2026-09-03-baseline-using-git-worktrees-pair.json +133 -0
- package/behavioral-evals/results/2026-09-03-summary-both.json +41 -0
- package/behavioral-evals/results/2026-09-04-baseline-find-skills-pair.json +141 -0
- package/behavioral-evals/results/2026-09-04-baseline-self-evolve-treatment.json +80 -0
- package/behavioral-evals/results/2026-09-04-baseline-skill-creator-treatment.json +80 -0
- package/behavioral-evals/results/2026-09-04-summary-both.json +41 -0
- package/behavioral-evals/run-with-plugin.js +15 -31
- package/behavioral-evals/run.js +446 -105
- package/behavioral-evals/transcript-parser.js +591 -0
- package/bin/cli.js +24 -12
- package/ci/ab-test-cases/grill-me-adversarial.yaml +14 -4
- package/ci/ab-test-cases/tdd-test-first.yaml +13 -6
- package/ci/ab-test-cases/verification-loop-gates.yaml +20 -7
- package/ci/ab-test-cases/verify-before-claim-cites.yaml +7 -6
- package/ci/ab-test-harness.js +113 -38
- package/ci/action-gate-platform-doc.test.js +31 -0
- package/ci/action-gate.test.js +260 -0
- package/ci/cases/case2.json +2 -2
- package/ci/consistency-check.js +51 -5
- package/ci/description-collision.js +15 -6
- package/ci/disclosure-check.js +225 -0
- package/ci/doc-capability-consistency.test.js +180 -0
- package/ci/fixtures/phase4/parallel-disjoint.stages.json +32 -0
- package/ci/fixtures/phase4/parallel-read-only.stages.json +32 -0
- package/ci/installer-roundtrip.test.js +159 -0
- package/ci/invariant-routing.test.js +74 -0
- package/ci/issue99-runtime.test.js +218 -0
- package/ci/mechanism-2aa-codex-action-gate-policy.test.js +117 -0
- package/ci/mechanism-2ab-openai-plugin-sync-integrity.test.js +80 -0
- package/ci/mechanism-2j-behavioral-runner.test.js +43 -1
- package/ci/mechanism-2k-reference-check.test.js +57 -2
- package/ci/mechanism-2l-multi-agent-workspace.test.js +202 -6
- package/ci/mechanism-2m-self-repo-guard.test.js +65 -0
- package/ci/mechanism-2n-opencode-plugin.test.js +461 -0
- package/ci/mechanism-2o-state-home-isolation.test.js +183 -0
- package/ci/mechanism-2p-installer-links.test.js +183 -0
- package/ci/mechanism-2q-installer-global-shared-store.test.js +132 -0
- package/ci/mechanism-2r-uninstall-flags.test.js +148 -0
- package/ci/mechanism-2s-tdd-quality.test.js +203 -0
- package/ci/mechanism-2t-disclosure.test.js +90 -0
- package/ci/mechanism-2t-evidence-triage.test.js +200 -0
- package/ci/mechanism-2t-issue-20.test.js +75 -0
- package/ci/mechanism-2t-platform-skill-targets.test.js +164 -0
- package/ci/mechanism-2t-transcript-parser.test.js +179 -0
- package/ci/mechanism-2t-workspace-identity.test.js +200 -0
- package/ci/mechanism-2u-issue-42-regressions.test.js +239 -0
- package/ci/mechanism-2u-issue-44.test.js +35 -0
- package/ci/mechanism-2v-verifier-boundaries.test.js +138 -0
- package/ci/mechanism-2w-platform-compatibility.test.js +148 -0
- package/ci/mechanism-2x-release-automation.test.js +157 -0
- package/ci/mechanism-2x-release-commit-message.test.js +66 -0
- package/ci/mechanism-2y-release-tooling-compat.test.js +47 -0
- package/ci/mechanism-2z-claude-plugin-manifest.test.js +29 -0
- package/ci/mutation-check.js +53 -0
- package/ci/mutation-loader.js +57 -0
- package/ci/negative-controls.js +76 -46
- package/ci/openai-plugin-package.test.js +182 -0
- package/ci/openai-plugin-runtime.test.js +175 -0
- package/ci/openai-plugin-submission.test.js +101 -0
- package/ci/phase4-ensemble-memory.test.js +307 -0
- package/ci/reference-check.js +73 -25
- package/ci/release-consistency-check.js +24 -1
- package/ci/router-phase2-safety.test.js +81 -0
- package/ci/router-workflow-plan.test.js +264 -0
- package/ci/routing-check.js +6 -5
- package/ci/runner.js +6 -5
- package/ci/skill-routing-check.js +48 -11
- package/ci/test-behavioral-case-validation.js +62 -0
- package/ci/test-helper.js +10 -0
- package/ci/workflow-plan-consumers-edge.test.js +72 -0
- package/ci/workflow-plan-consumers.test.js +276 -0
- package/docs/architecture.md +137 -57
- package/docs/audit.md +22 -20
- package/docs/codex-action-gate.md +90 -0
- package/docs/issue-20-rollup.md +47 -0
- package/docs/mechanism-first-skill-mesh.md +22 -20
- package/docs/openai-plugin.md +118 -0
- package/docs/philosophy.md +28 -17
- package/docs/platform-capabilities.md +73 -0
- package/docs/platform-compatibility.json +270 -0
- package/docs/reflection.md +37 -22
- package/docs/release-evidence.md +25 -0
- package/docs/routing.md +157 -44
- package/docs/security-audit.md +38 -0
- package/docs/troubleshooting.md +82 -46
- package/docs/verifier-boundaries.md +45 -0
- package/docs/workflows/fable-mode.md +77 -21
- package/docs/workflows/find-skills.md +40 -0
- package/docs/workflows/grill-with-docs.md +2 -2
- package/docs/workflows/multi-agent-workspace.md +22 -10
- package/docs/workflows/self-evolve.md +33 -0
- package/docs/workflows/tdd.md +36 -0
- package/docs/workflows/todo-driven-workflow.md +33 -0
- package/environment-detection/SKILL.md +3 -3
- package/eval-harness/SKILL.md +7 -7
- package/eval-harness/scripts/evaluate.js +12 -10
- package/evals/evaluation-reports.md +14 -0
- package/evals/harness-everything/tasks/positive-tier1.yaml +1 -1
- package/evals/harness-everything/tasks/positive-tier2.yaml +1 -1
- package/evals/harness-everything/tasks/positive-tier3.yaml +1 -1
- package/evals/install-cognitive-os/tasks/positive-trigger.yaml +1 -1
- package/fable-discipline/SKILL.md +2 -2
- package/fable-mode/CONTRACT-FORMAT.md +147 -33
- package/fable-mode/SKILL.md +3 -3
- package/fable-mode/agents/fable-orchestrator.md +95 -35
- package/fable-mode/execution-guardrails/SKILL.md +2 -2
- package/fable-mode/fable-haiku/SKILL.md +3 -3
- package/fable-mode/fable-opus/SKILL.md +3 -3
- package/fable-mode/fable-sonnet/SKILL.md +3 -3
- package/fable-mode/scripts/ensemble-review.js +371 -0
- package/fable-mode/scripts/model-selector.js +134 -3
- package/fable-mode/scripts/workflow-plan-consumer.js +337 -0
- package/find-skills/SKILL.md +5 -5
- package/find-skills/references/discovery-flow.md +1 -1
- package/find-skills/scripts/use-skill.js +7 -1
- package/git-commit/SKILL.md +4 -11
- package/grill-me/SKILL.md +5 -5
- package/grill-with-docs/SKILL.md +7 -7
- package/grill-with-docs/references/session-playbook.md +6 -4
- package/harness-everything/SKILL.md +23 -22
- package/harness-everything/references/router-workflow-plan.md +128 -0
- package/harness-everything/references/triage-and-tiers.md +173 -96
- package/harness-everything/schemas/router-task-shape.schema.json +126 -0
- package/harness-everything/schemas/router-workflow-plan.schema.json +149 -0
- package/harness-everything/scripts/bootstrap.js +13 -4
- package/harness-everything/scripts/ensemble-policy.js +102 -0
- package/harness-everything/scripts/kernel-router.js +212 -0
- package/harness-everything/scripts/router-contract.js +612 -0
- package/harness-everything/scripts/self-heal.js +29 -20
- package/harness-everything/scripts/tier-router.js +346 -221
- package/hooks/hooks.json +60 -15
- package/hooks/scripts/action-gate-rules.json +77 -0
- package/hooks/scripts/action-gate.js +528 -0
- package/hooks/scripts/atomic-commit-check.js +5 -5
- package/hooks/scripts/codex-action-gate-post.js +46 -0
- package/hooks/scripts/codex-permission-request.js +140 -0
- package/hooks/scripts/context-compact.js +4 -3
- package/hooks/scripts/contract-test.js +118 -51
- package/hooks/scripts/lib/fable-contracts.js +121 -0
- package/hooks/scripts/lib/harness-state.js +164 -28
- package/hooks/scripts/lib/platforms/claude.js +45 -62
- package/hooks/scripts/lib/platforms/codex.js +35 -61
- package/hooks/scripts/lib/platforms/continue.js +7 -7
- package/hooks/scripts/lib/platforms/copilot.js +29 -53
- package/hooks/scripts/lib/platforms/hermes.js +71 -20
- package/hooks/scripts/openai-session-start.js +81 -0
- package/hooks/scripts/rule-of-3-tracker.js +9 -8
- package/hooks/scripts/rule-of-3.js +29 -21
- package/hooks/scripts/state-persist.js +5 -4
- package/hooks/scripts/stop-gate.js +2 -2
- package/hooks/scripts/subagent-scope-guard.js +144 -38
- package/hooks/scripts/subagent-start.js +19 -0
- package/hooks/scripts/subagent-stop.js +6 -0
- package/improve-codebase-architecture/SKILL.md +2 -4
- package/improve-codebase-architecture/references/README.md +16 -0
- package/install-cognitive-os/SKILL.md +28 -22
- package/multi-agent-workspace/SKILL.md +16 -15
- package/multi-agent-workspace/references/architecture-guide.md +13 -9
- package/multi-agent-workspace/references/orchestration.md +39 -22
- package/multi-agent-workspace/scripts/consume-workflow-plan.js +129 -0
- package/multi-agent-workspace/scripts/index_memory.js +79 -26
- package/multi-agent-workspace/scripts/path-boundary.js +59 -0
- package/multi-agent-workspace/scripts/project-docs-resolver.js +357 -0
- package/multi-agent-workspace/scripts/scaffold.js +258 -103
- package/multi-agent-workspace/templates/AGENTS.md +20 -0
- package/multi-agent-workspace/workflows/03-generation.md +8 -10
- package/multi-agent-workspace/workflows/04-launcher.md +5 -5
- package/opencode-plugin/README.md +114 -71
- package/opencode-plugin/hooks/verify.js +62 -12
- package/opencode-plugin/index.mjs +547 -0
- package/opencode-plugin/plugin.json +4 -12
- package/package.json +16 -3
- package/plugins/harness-everything/.codex-plugin/plugin.json +39 -0
- package/plugins/harness-everything/hooks/hooks.json +197 -0
- package/plugins/harness-everything/hooks/scripts/action-gate-rules.json +77 -0
- package/plugins/harness-everything/hooks/scripts/action-gate.js +528 -0
- package/plugins/harness-everything/hooks/scripts/atomic-commit-check.js +68 -0
- package/plugins/harness-everything/hooks/scripts/boundary-guard.js +55 -0
- package/plugins/harness-everything/hooks/scripts/codex-action-gate-post.js +46 -0
- package/plugins/harness-everything/hooks/scripts/codex-permission-request.js +140 -0
- package/plugins/harness-everything/hooks/scripts/context-compact.js +115 -0
- package/plugins/harness-everything/hooks/scripts/contract-test.js +146 -0
- package/plugins/harness-everything/hooks/scripts/depth-guard.js +65 -0
- package/plugins/harness-everything/hooks/scripts/lib/fable-contracts.js +121 -0
- package/plugins/harness-everything/hooks/scripts/lib/harness-state.js +275 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/claude.js +168 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/codex.js +88 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/continue.js +91 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/copilot.js +85 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/cursor.js +89 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/hermes.js +108 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/index.js +17 -0
- package/plugins/harness-everything/hooks/scripts/lib/platforms/worktrees.js +20 -0
- package/plugins/harness-everything/hooks/scripts/lib/workspace.js +290 -0
- package/plugins/harness-everything/hooks/scripts/openai-session-start.js +81 -0
- package/plugins/harness-everything/hooks/scripts/reset-circuit-breaker.js +30 -0
- package/plugins/harness-everything/hooks/scripts/rule-of-3-tracker.js +125 -0
- package/plugins/harness-everything/hooks/scripts/rule-of-3.js +195 -0
- package/plugins/harness-everything/hooks/scripts/state-persist.js +100 -0
- package/plugins/harness-everything/hooks/scripts/stop-gate.js +80 -0
- package/plugins/harness-everything/hooks/scripts/subagent-scope-guard.js +194 -0
- package/plugins/harness-everything/hooks/scripts/subagent-start.js +19 -0
- package/plugins/harness-everything/hooks/scripts/subagent-stop.js +6 -0
- package/plugins/harness-everything/hooks/session-start.js +81 -0
- package/plugins/harness-everything/plugin.json +43 -0
- package/plugins/harness-everything/scripts/lib/advisory-text.js +174 -0
- package/plugins/harness-everything/skills/environment-detection/SKILL.md +41 -0
- package/plugins/harness-everything/skills/environment-detection/references/shell-syntax-rules.md +76 -0
- package/plugins/harness-everything/skills/environment-detection/scripts/preflight.js +103 -0
- package/plugins/harness-everything/skills/eval-harness/SKILL.md +42 -0
- package/plugins/harness-everything/skills/eval-harness/references/scoring-rubric.md +68 -0
- package/plugins/harness-everything/skills/eval-harness/scripts/evaluate.js +79 -0
- package/plugins/harness-everything/skills/fable-discipline/SKILL.md +41 -0
- package/plugins/harness-everything/skills/fable-discipline/references/discipline-rules.md +22 -0
- package/plugins/harness-everything/skills/fable-mode/CONTRACT-FORMAT.md +165 -0
- package/plugins/harness-everything/skills/fable-mode/SKILL.md +39 -0
- package/plugins/harness-everything/skills/fable-mode/agents/fable-orchestrator.md +159 -0
- package/plugins/harness-everything/skills/fable-mode/agents/fable-verifier.md +32 -0
- package/plugins/harness-everything/skills/fable-mode/agents/fable-worker-haiku.md +52 -0
- package/plugins/harness-everything/skills/fable-mode/agents/fable-worker-sonnet.md +53 -0
- package/plugins/harness-everything/skills/fable-mode/execution-guardrails/SKILL.md +43 -0
- package/plugins/harness-everything/skills/fable-mode/execution-guardrails/references/guardrail-rules.md +56 -0
- package/plugins/harness-everything/skills/fable-mode/fable-haiku/SKILL.md +36 -0
- package/plugins/harness-everything/skills/fable-mode/fable-haiku/references/routing-notes.md +26 -0
- package/plugins/harness-everything/skills/fable-mode/fable-opus/SKILL.md +36 -0
- package/plugins/harness-everything/skills/fable-mode/fable-sonnet/SKILL.md +36 -0
- package/plugins/harness-everything/skills/fable-mode/model-matrix.json +32 -0
- package/plugins/harness-everything/skills/fable-mode/references/execution-phases.md +37 -0
- package/plugins/harness-everything/skills/fable-mode/references/model-matrix.md +39 -0
- package/plugins/harness-everything/skills/fable-mode/scripts/ensemble-review.js +371 -0
- package/plugins/harness-everything/skills/fable-mode/scripts/model-selector.js +260 -0
- package/plugins/harness-everything/skills/fable-mode/scripts/workflow-plan-consumer.js +337 -0
- package/plugins/harness-everything/skills/find-skills/SKILL.md +48 -0
- package/plugins/harness-everything/skills/find-skills/references/discovery-flow.md +102 -0
- package/plugins/harness-everything/skills/find-skills/scripts/use-skill.js +112 -0
- package/plugins/harness-everything/skills/git-commit/SKILL.md +38 -0
- package/plugins/harness-everything/skills/git-commit/guides/ANGULAR_STYLE.md +61 -0
- package/plugins/harness-everything/skills/git-commit/guides/COMMIT_GENERATION.md +24 -0
- package/plugins/harness-everything/skills/git-commit/guides/LANGUAGE_DETECTION.md +16 -0
- package/plugins/harness-everything/skills/git-commit/guides/MAIN_REPO.md +15 -0
- package/plugins/harness-everything/skills/git-commit/guides/SUBMODULES.md +27 -0
- package/plugins/harness-everything/skills/git-commit/references/commit-flow.md +30 -0
- package/plugins/harness-everything/skills/grill-me/SKILL.md +43 -0
- package/plugins/harness-everything/skills/grill-me/references/grilling-playbook.md +38 -0
- package/plugins/harness-everything/skills/grill-with-docs/ADR-FORMAT.md +47 -0
- package/plugins/harness-everything/skills/grill-with-docs/CONTEXT-FORMAT.md +63 -0
- package/plugins/harness-everything/skills/grill-with-docs/SKILL.md +41 -0
- package/plugins/harness-everything/skills/grill-with-docs/references/session-playbook.md +99 -0
- package/plugins/harness-everything/skills/harness-everything/SKILL.md +43 -0
- package/plugins/harness-everything/skills/harness-everything/examples/README.md +9 -0
- package/plugins/harness-everything/skills/harness-everything/examples/todo-workflow.md +22 -0
- package/plugins/harness-everything/skills/harness-everything/references/router-workflow-plan.md +128 -0
- package/plugins/harness-everything/skills/harness-everything/references/skill-registry.md +39 -0
- package/plugins/harness-everything/skills/harness-everything/references/triage-and-tiers.md +193 -0
- package/plugins/harness-everything/skills/harness-everything/schemas/router-task-shape.schema.json +126 -0
- package/plugins/harness-everything/skills/harness-everything/schemas/router-workflow-plan.schema.json +149 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/bootstrap.js +212 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/ensemble-policy.js +102 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/kernel-router.js +212 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/router-contract.js +612 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/routing-keywords.json +320 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/self-heal.js +250 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/tier-router.js +459 -0
- package/plugins/harness-everything/skills/harness-everything/scripts/verify-gate.js +100 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/SKILL.md +38 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/guides/DEEPENING.md +37 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/guides/HTML-REPORT.md +123 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/guides/INTERFACE-DESIGN.md +44 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/guides/LANGUAGE.md +53 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/guides/refactoring-workflow.md +59 -0
- package/plugins/harness-everything/skills/improve-codebase-architecture/references/README.md +16 -0
- package/plugins/harness-everything/skills/install-cognitive-os/SKILL.md +44 -0
- package/plugins/harness-everything/skills/install-cognitive-os/references/cognitive-loop.md +27 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/NOTICE-agency-agents.md +27 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/SKILL.md +38 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/examples/csharp-monorepo.md +123 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/examples/typescript-cli.md +126 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-claude.md +77 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-codex.md +146 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-continue.md +38 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-copilot.md +74 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-cursor.md +40 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/platform-hermes.md +38 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/guidelines/role-capabilities.md +108 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/references/agency-agents.md +31 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/references/architecture-guide.md +36 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/references/orchestration.md +65 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/agency-catalog.js +190 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/consume-workflow-plan.js +129 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/index_memory.js +104 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/path-boundary.js +59 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/project-docs-resolver.js +357 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/scripts/scaffold.js +408 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/AGENTS.md +20 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/README.md +17 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/agent-launcher.md +68 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/backend-developer-csharp.md +56 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/backend-developer-polyglot.md +62 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/backend-developer-ts.md +57 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/memory-keeper.md +37 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/requirement-analyzer.md +66 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/specialist-agent.md +34 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/task-verifier.md +56 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/templates/tool-maker.md +33 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/workflows/01-init.md +22 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/workflows/02-analysis.md +23 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/workflows/03-generation.md +25 -0
- package/plugins/harness-everything/skills/multi-agent-workspace/workflows/04-launcher.md +16 -0
- package/plugins/harness-everything/skills/repo-docs/SKILL.md +42 -0
- package/plugins/harness-everything/skills/repo-docs/references/process-guide.md +60 -0
- package/plugins/harness-everything/skills/repo-docs/templates/agents-template.md +264 -0
- package/plugins/harness-everything/skills/repo-docs/templates/knowledge-agents-template.md +50 -0
- package/plugins/harness-everything/skills/repo-docs/templates/knowledge-readme-template.md +83 -0
- package/plugins/harness-everything/skills/repo-docs/templates/manual-recon.md +117 -0
- package/plugins/harness-everything/skills/repo-docs/templates/multi-skills-readme-template.md +72 -0
- package/plugins/harness-everything/skills/repo-docs/templates/product-agents-template.md +78 -0
- package/plugins/harness-everything/skills/repo-docs/templates/product-readme-template.md +95 -0
- package/plugins/harness-everything/skills/repo-docs/templates/readme-template.md +136 -0
- package/plugins/harness-everything/skills/repo-docs/templates/single-skill-readme-template.md +70 -0
- package/plugins/harness-everything/skills/repo-docs/templates/skills-agents-template.md +60 -0
- package/plugins/harness-everything/skills/rewrite-commits/SKILL.md +37 -0
- package/plugins/harness-everything/skills/rewrite-commits/references/workflow.md +18 -0
- package/plugins/harness-everything/skills/security-review/SKILL.md +47 -0
- package/plugins/harness-everything/skills/security-review/guides/OWASP-PATTERNS.md +31 -0
- package/plugins/harness-everything/skills/security-review/guides/STRIDE-THREAT-MODEL.md +21 -0
- package/plugins/harness-everything/skills/security-review/references/security-checklist.md +478 -0
- package/plugins/harness-everything/skills/security-review/scripts/audit-secrets.js +85 -0
- package/plugins/harness-everything/skills/self-evolve/SKILL.md +45 -0
- package/plugins/harness-everything/skills/self-evolve/references/memory-resolution.md +61 -0
- package/plugins/harness-everything/skills/self-evolve/scripts/persist-memory.js +272 -0
- package/plugins/harness-everything/skills/self-evolve/scripts/register-dynamic-skill.js +253 -0
- package/plugins/harness-everything/skills/self-evolve/scripts/self-regression.js +125 -0
- package/plugins/harness-everything/skills/skill-creator/SKILL.md +47 -0
- package/plugins/harness-everything/skills/skill-creator/references/authoring-workflow.md +14 -0
- package/plugins/harness-everything/skills/skill-creator/references/dynamic-generation-contract.md +21 -0
- package/plugins/harness-everything/skills/skill-creator/references/quality-checklist.md +16 -0
- package/plugins/harness-everything/skills/skill-creator/references/quality-principles.md +81 -0
- package/plugins/harness-everything/skills/skill-creator/references/testing-workflow.md +27 -0
- package/plugins/harness-everything/skills/skill-style/SKILL.md +42 -0
- package/plugins/harness-everything/skills/skill-style/references/style-guide.md +62 -0
- package/plugins/harness-everything/skills/tdd/SKILL.md +38 -0
- package/plugins/harness-everything/skills/tdd/fixtures/pass-evidence.json +100 -0
- package/plugins/harness-everything/skills/tdd/fixtures/quality-scenarios.json +20 -0
- package/plugins/harness-everything/skills/tdd/guides/deep-modules.md +33 -0
- package/plugins/harness-everything/skills/tdd/guides/interface-design.md +31 -0
- package/plugins/harness-everything/skills/tdd/guides/mocking.md +59 -0
- package/plugins/harness-everything/skills/tdd/guides/refactoring.md +10 -0
- package/plugins/harness-everything/skills/tdd/guides/tests.md +61 -0
- package/plugins/harness-everything/skills/tdd/references/core-discipline.md +51 -0
- package/plugins/harness-everything/skills/tdd/references/integration-testing.md +13 -0
- package/plugins/harness-everything/skills/tdd/references/quality-model.md +68 -0
- package/plugins/harness-everything/skills/tdd/references/unit-testing.md +15 -0
- package/plugins/harness-everything/skills/tdd/schemas/quality-evidence.schema.json +211 -0
- package/plugins/harness-everything/skills/tdd/schemas/quality-report.schema.json +110 -0
- package/plugins/harness-everything/skills/tdd/scripts/quality-gate.js +475 -0
- package/plugins/harness-everything/skills/to-spec/SKILL.md +40 -0
- package/plugins/harness-everything/skills/to-spec/references/process.md +77 -0
- package/plugins/harness-everything/skills/to-spec/scripts/check-project-docs.js +239 -0
- package/plugins/harness-everything/skills/to-spec/scripts/project-docs-resolver.js +111 -0
- package/plugins/harness-everything/skills/to-spec/templates/cli-reference.md +31 -0
- package/plugins/harness-everything/skills/to-spec/templates/dev-doc.md +27 -0
- package/plugins/harness-everything/skills/to-spec/templates/feature-spec.md +59 -0
- package/plugins/harness-everything/skills/to-spec/templates/schema-doc.md +31 -0
- package/plugins/harness-everything/skills/to-tickets/SKILL.md +42 -0
- package/plugins/harness-everything/skills/to-tickets/references/publishing-and-templates.md +97 -0
- package/plugins/harness-everything/skills/todo-driven-workflow/SKILL.md +42 -0
- package/plugins/harness-everything/skills/todo-driven-workflow/references/execution-guide.md +27 -0
- package/plugins/harness-everything/skills/using-git-worktrees/SKILL.md +49 -0
- package/plugins/harness-everything/skills/using-git-worktrees/references/workflow-details.md +108 -0
- package/plugins/harness-everything/skills/verification-loop/SKILL.md +46 -0
- package/plugins/harness-everything/skills/verification-loop/references/verification-phases.md +116 -0
- package/plugins/harness-everything/skills/verification-loop/templates/verification-report.template.md +15 -0
- package/plugins/harness-everything/skills/verify-before-claim/SKILL.md +43 -0
- package/plugins/harness-everything/skills/verify-before-claim/references/verification-guide.md +73 -0
- package/plugins/harness-everything/skills/zoom-out/SKILL.md +42 -0
- package/plugins/harness-everything/skills/zoom-out/references/circuit-breaker.md +66 -0
- package/plugins/harness-everything/skills/zoom-out/templates/zoom-out-report.template.md +20 -0
- package/repo-docs/SKILL.md +3 -7
- package/rewrite-commits/SKILL.md +2 -2
- package/scripts/build-openai-submission.js +158 -0
- package/scripts/installer.js +86 -11
- package/scripts/lib/advisory-text.js +25 -18
- package/scripts/lib/execution-contract.js +70 -0
- package/scripts/lib/manifest.js +34 -38
- package/scripts/lib/path-boundary.js +90 -0
- package/scripts/lib/skills.js +124 -70
- package/scripts/lib/workspace.js +269 -7
- package/scripts/sync-openai-plugin.js +91 -0
- package/scripts/sync-release-version.js +212 -0
- package/security-review/SKILL.md +4 -7
- package/security-review/scripts/audit-secrets.js +9 -8
- package/self-evolve/SKILL.md +9 -11
- package/self-evolve/references/memory-resolution.md +4 -1
- package/self-evolve/scripts/persist-memory.js +221 -88
- package/self-evolve/scripts/register-dynamic-skill.js +12 -10
- package/self-evolve/scripts/self-regression.js +23 -8
- package/skill-creator/SKILL.md +12 -14
- package/skill-creator/references/quality-checklist.md +1 -0
- package/skill-style/SKILL.md +4 -4
- package/skill-style/references/style-guide.md +29 -3
- package/submission/openai/README.md +67 -0
- package/submission/openai/listing.json +31 -0
- package/submission/openai/test-cases.json +61 -0
- package/tdd/SKILL.md +20 -26
- package/tdd/fixtures/pass-evidence.json +100 -0
- package/tdd/fixtures/quality-scenarios.json +20 -0
- package/tdd/references/core-discipline.md +2 -0
- package/tdd/references/integration-testing.md +13 -0
- package/tdd/references/quality-model.md +68 -0
- package/tdd/references/unit-testing.md +15 -0
- package/tdd/schemas/quality-evidence.schema.json +211 -0
- package/tdd/schemas/quality-report.schema.json +110 -0
- package/tdd/scripts/quality-gate.js +475 -0
- package/to-spec/SKILL.md +3 -5
- package/to-spec/scripts/check-project-docs.js +42 -62
- package/to-spec/scripts/project-docs-resolver.js +111 -0
- package/to-tickets/SKILL.md +4 -4
- package/todo-driven-workflow/SKILL.md +6 -4
- package/using-git-worktrees/SKILL.md +3 -3
- package/verification-loop/SKILL.md +4 -4
- package/verify-before-claim/SKILL.md +2 -2
- package/zoom-out/SKILL.md +4 -4
- package/ab-test-deep-analysis-2026-08-27.md +0 -383
- package/ab-test-results-2026-08-27.md +0 -152
- package/ab-test-statistical-analysis-31cases.md +0 -193
- package/opencode-plugin/hooks/circuit-breaker.js +0 -121
- package/opencode-plugin/hooks/compliance.js +0 -128
- package/opencode-plugin/hooks/post-edit.js +0 -62
- package/opencode-plugin/hooks/pre-complete.js +0 -65
- package/opencode-plugin/test-hook-firing.js +0 -198
|
@@ -0,0 +1,547 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Harness Enforcement Plugin for opencode
|
|
3
|
+
*
|
|
4
|
+
* A single self-contained ESM module, because that is what opencode actually
|
|
5
|
+
* loads: a JS/TS file dropped in `.opencode/plugins/` (or referenced from
|
|
6
|
+
* `opencode.json`) that exports a function returning a hooks object. There is
|
|
7
|
+
* no JSON-manifest-to-script mechanism - the old plugin.json + hooks/*.js
|
|
8
|
+
* split assumed one and was never invoked by opencode. See issue #37.
|
|
9
|
+
*
|
|
10
|
+
* Hook mapping (old CLI script -> real opencode hook):
|
|
11
|
+
* post-edit.js -> tool.execute.after (fires once per edit/write/apply_patch)
|
|
12
|
+
* pre-complete.js -> event (session.idle) (opencode has no "before complete"
|
|
13
|
+
* hook; session.idle is the closest analog - the agent's
|
|
14
|
+
* turn just ended. There is no way to *block* that event,
|
|
15
|
+
* so pending verification is enforced by running it right
|
|
16
|
+
* there and, on failure, pushing a synthetic follow-up
|
|
17
|
+
* message via client.session.prompt() to force the agent
|
|
18
|
+
* to keep going instead of truly stopping.)
|
|
19
|
+
* verify.js -> inlined into the session.idle handler
|
|
20
|
+
* circuit-breaker.js -> tool.execute.before (throws to hard-block edit tools
|
|
21
|
+
* once locked) + failure tracking inside session.idle
|
|
22
|
+
* compliance.js -> folded into the same state writes, no separate stage
|
|
23
|
+
*
|
|
24
|
+
* The circuit breaker trips on repeated *verification* failures (same failing
|
|
25
|
+
* command + truncated error), not on arbitrary tool failures: tool.execute.after
|
|
26
|
+
* gives back `{title, output, metadata}` for every tool with no normalized
|
|
27
|
+
* success/failure field, so there is no reliable cross-tool failure signal to
|
|
28
|
+
* key a signature on. Verification failures are self-produced and already
|
|
29
|
+
* normalized, so that is the honest signal to use.
|
|
30
|
+
*/
|
|
31
|
+
|
|
32
|
+
import { homedir } from "node:os"
|
|
33
|
+
import { join, dirname, resolve, basename, relative, isAbsolute } from "node:path"
|
|
34
|
+
import { readFileSync, writeFileSync, existsSync, mkdirSync, rmSync, realpathSync, readdirSync } from "node:fs"
|
|
35
|
+
import { execSync } from "node:child_process"
|
|
36
|
+
import { createHash, randomUUID } from "node:crypto"
|
|
37
|
+
|
|
38
|
+
const EDIT_TOOLS = new Set(["edit", "write", "apply_patch"])
|
|
39
|
+
|
|
40
|
+
// Global state root, matching the Node-side resolver at
|
|
41
|
+
// scripts/lib/workspace.js#getStateHome (same env var precedence) - this
|
|
42
|
+
// plugin can't require that CJS module (opencode loads a single self-
|
|
43
|
+
// contained ESM file, see the header comment above), so the derivation is
|
|
44
|
+
// duplicated here, algorithm-for-algorithm, so the two stay physically
|
|
45
|
+
// compatible if a workspace is ever inspected from both sides.
|
|
46
|
+
function getStateHome() {
|
|
47
|
+
return process.env.HARNESS_STATE_HOME || join(homedir(), ".agents", "harness-everything")
|
|
48
|
+
}
|
|
49
|
+
|
|
50
|
+
// Keys state per real workspace (`directory`, which opencode itself resolves
|
|
51
|
+
// - never a cwd walk) instead of one flat dir shared by every project on the
|
|
52
|
+
// machine. The pre-fix `~/.harness-state` had no such key at all: a circuit
|
|
53
|
+
// breaker trip in one opencode project hard-locked every other one too
|
|
54
|
+
// (issue #42 item #4).
|
|
55
|
+
function getWorkspaceKey(directory) {
|
|
56
|
+
let real = resolve(directory)
|
|
57
|
+
try { real = realpathSync(real) } catch { /* directory may not exist yet */ }
|
|
58
|
+
const slug = basename(real).toLowerCase().replace(/[^a-z0-9._-]+/g, "-").replace(/^-+|-+$/g, "") || "workspace"
|
|
59
|
+
const hashInput = process.platform === "win32" ? real.toLowerCase() : real
|
|
60
|
+
const hash = createHash("sha1").update(hashInput).digest("hex").slice(0, 12)
|
|
61
|
+
return `${slug}-${hash}`
|
|
62
|
+
}
|
|
63
|
+
|
|
64
|
+
// Move the pre-#42 flat plugin state into the first workspace that claims it.
|
|
65
|
+
// Keep conflicting or unsupported entries in place so a later run can retry.
|
|
66
|
+
function migrateLegacyFlatState(stateDir) {
|
|
67
|
+
const legacyDir = join(homedir(), ".harness-state")
|
|
68
|
+
if (!existsSync(legacyDir)) return
|
|
69
|
+
try {
|
|
70
|
+
mkdirSync(stateDir, { recursive: true })
|
|
71
|
+
let conflict = false
|
|
72
|
+
for (const entry of readdirSync(legacyDir, { withFileTypes: true })) {
|
|
73
|
+
if (!entry.isFile()) {
|
|
74
|
+
conflict = true
|
|
75
|
+
continue
|
|
76
|
+
}
|
|
77
|
+
const source = join(legacyDir, entry.name)
|
|
78
|
+
const destination = join(stateDir, entry.name)
|
|
79
|
+
if (!existsSync(destination)) writeFileSync(destination, readFileSync(source))
|
|
80
|
+
else if (!readFileSync(source).equals(readFileSync(destination))) conflict = true
|
|
81
|
+
}
|
|
82
|
+
if (!conflict) rmSync(legacyDir, { recursive: true, force: true })
|
|
83
|
+
} catch {
|
|
84
|
+
// Best effort: leave the legacy source available for recovery.
|
|
85
|
+
}
|
|
86
|
+
}
|
|
87
|
+
|
|
88
|
+
function getStateDir(directory) {
|
|
89
|
+
return join(getStateHome(), "workspaces", getWorkspaceKey(directory))
|
|
90
|
+
}
|
|
91
|
+
|
|
92
|
+
function getStateRoot(directory) {
|
|
93
|
+
return join(getStateDir(directory), "state")
|
|
94
|
+
}
|
|
95
|
+
|
|
96
|
+
function getSessionKey(sessionID) {
|
|
97
|
+
const id = String(sessionID || "default")
|
|
98
|
+
const windowsReserved = /^(?:con|prn|aux|nul|com[1-9]|lpt[1-9])(?:\..*)?$/i
|
|
99
|
+
if (
|
|
100
|
+
/^[A-Za-z0-9._-]+$/.test(id) &&
|
|
101
|
+
id !== "." &&
|
|
102
|
+
id !== ".." &&
|
|
103
|
+
id === id.trim() &&
|
|
104
|
+
!/[. ]$/.test(id) &&
|
|
105
|
+
!windowsReserved.test(id)
|
|
106
|
+
) return id
|
|
107
|
+
return `session-${createHash("sha1").update(id).digest("hex").slice(0, 12)}`
|
|
108
|
+
}
|
|
109
|
+
|
|
110
|
+
function isWithin(parent, candidate) {
|
|
111
|
+
const relativePath = relative(resolve(parent), resolve(candidate))
|
|
112
|
+
return relativePath === "" || (
|
|
113
|
+
relativePath !== ".." &&
|
|
114
|
+
!relativePath.startsWith(`..${process.platform === "win32" ? "\\" : "/"}`) &&
|
|
115
|
+
!isAbsolute(relativePath)
|
|
116
|
+
)
|
|
117
|
+
}
|
|
118
|
+
|
|
119
|
+
function samePath(left, right) {
|
|
120
|
+
const a = resolve(left)
|
|
121
|
+
const b = resolve(right)
|
|
122
|
+
return process.platform === "win32" ? a.toLowerCase() === b.toLowerCase() : a === b
|
|
123
|
+
}
|
|
124
|
+
|
|
125
|
+
function getSessionDir(directory, sessionID) {
|
|
126
|
+
const sessionsRoot = resolve(join(getStateRoot(directory), "sessions"))
|
|
127
|
+
const candidate = resolve(sessionsRoot, getSessionKey(sessionID))
|
|
128
|
+
if (!isWithin(sessionsRoot, candidate) || candidate === sessionsRoot) {
|
|
129
|
+
const fallback = `session-${createHash("sha1").update(String(sessionID || "default")).digest("hex").slice(0, 12)}`
|
|
130
|
+
return join(sessionsRoot, fallback)
|
|
131
|
+
}
|
|
132
|
+
return candidate
|
|
133
|
+
}
|
|
134
|
+
|
|
135
|
+
function loadJSON(file, fallback) {
|
|
136
|
+
try {
|
|
137
|
+
if (existsSync(file)) return JSON.parse(readFileSync(file, "utf8"))
|
|
138
|
+
} catch {
|
|
139
|
+
// fall through to default
|
|
140
|
+
}
|
|
141
|
+
return fallback
|
|
142
|
+
}
|
|
143
|
+
|
|
144
|
+
function saveJSON(file, data) {
|
|
145
|
+
mkdirSync(dirname(file), { recursive: true })
|
|
146
|
+
writeFileSync(file, JSON.stringify(data, null, 2))
|
|
147
|
+
}
|
|
148
|
+
|
|
149
|
+
function defaultEditState() {
|
|
150
|
+
return {
|
|
151
|
+
lastEditTime: null,
|
|
152
|
+
verificationPending: false,
|
|
153
|
+
editsSinceVerification: 0,
|
|
154
|
+
followUpPending: false,
|
|
155
|
+
followUpDeliveryPending: false,
|
|
156
|
+
followUpMessage: null,
|
|
157
|
+
lastFollowUpError: null,
|
|
158
|
+
sessionStart: Date.now(),
|
|
159
|
+
}
|
|
160
|
+
}
|
|
161
|
+
|
|
162
|
+
function defaultBreakerState() {
|
|
163
|
+
return {
|
|
164
|
+
failures: {},
|
|
165
|
+
hardLock: false,
|
|
166
|
+
lastReflection: null,
|
|
167
|
+
lastReflectionSignature: null,
|
|
168
|
+
reflectionPending: false,
|
|
169
|
+
reflectionRequestedAt: null,
|
|
170
|
+
reflectionToken: null,
|
|
171
|
+
reflectionSignature: null,
|
|
172
|
+
}
|
|
173
|
+
}
|
|
174
|
+
|
|
175
|
+
function normalizeEditState(state) {
|
|
176
|
+
return { ...defaultEditState(), ...(state || {}) }
|
|
177
|
+
}
|
|
178
|
+
|
|
179
|
+
function normalizeBreakerState(state) {
|
|
180
|
+
return { ...defaultBreakerState(), ...(state || {}), failures: (state && state.failures) || {} }
|
|
181
|
+
}
|
|
182
|
+
|
|
183
|
+
function isRecord(value) {
|
|
184
|
+
return value !== null && typeof value === "object" && !Array.isArray(value)
|
|
185
|
+
}
|
|
186
|
+
|
|
187
|
+
function loadBreakerState(file) {
|
|
188
|
+
if (!existsSync(file)) return defaultBreakerState()
|
|
189
|
+
try {
|
|
190
|
+
const state = JSON.parse(readFileSync(file, "utf8"))
|
|
191
|
+
const finiteOrNull = (value) => value === undefined || value === null || Number.isFinite(value)
|
|
192
|
+
const stringOrNull = (value) => value === undefined || value === null || typeof value === "string"
|
|
193
|
+
if (
|
|
194
|
+
!isRecord(state) ||
|
|
195
|
+
(state.failures !== undefined && !isRecord(state.failures)) ||
|
|
196
|
+
(state.hardLock !== undefined && typeof state.hardLock !== "boolean") ||
|
|
197
|
+
(state.reflectionPending !== undefined && typeof state.reflectionPending !== "boolean") ||
|
|
198
|
+
!finiteOrNull(state.lastReflection) ||
|
|
199
|
+
!finiteOrNull(state.reflectionRequestedAt) ||
|
|
200
|
+
!stringOrNull(state.lastReflectionSignature) ||
|
|
201
|
+
!stringOrNull(state.reflectionToken) ||
|
|
202
|
+
!stringOrNull(state.reflectionSignature) ||
|
|
203
|
+
Object.values(state.failures || {}).some((entry) =>
|
|
204
|
+
!isRecord(entry) ||
|
|
205
|
+
(entry.count !== undefined && (!Number.isInteger(entry.count) || entry.count < 0)) ||
|
|
206
|
+
!finiteOrNull(entry.firstSeen) ||
|
|
207
|
+
!finiteOrNull(entry.lastSeen)
|
|
208
|
+
)
|
|
209
|
+
) throw new Error("invalid breaker state shape")
|
|
210
|
+
return normalizeBreakerState(state)
|
|
211
|
+
} catch {
|
|
212
|
+
throw new Error(`Harness circuit breaker state is corrupt: "${file}" cannot be read safely.`)
|
|
213
|
+
}
|
|
214
|
+
}
|
|
215
|
+
|
|
216
|
+
const REFLECTION_FILE = "zoom-out-report.md"
|
|
217
|
+
const REFLECTION_SECTIONS = ["## Goal", "## Failed Attempts", "## Verified Facts", "## Diagnosis", "## Decision"]
|
|
218
|
+
|
|
219
|
+
function reflectionReportPath(stateDir) {
|
|
220
|
+
return join(stateDir, REFLECTION_FILE)
|
|
221
|
+
}
|
|
222
|
+
|
|
223
|
+
function isValidReflection(reportFile, breaker) {
|
|
224
|
+
if (!breaker.reflectionPending || !breaker.reflectionToken || !existsSync(reportFile)) return false
|
|
225
|
+
try {
|
|
226
|
+
const report = readFileSync(reportFile, "utf8")
|
|
227
|
+
const lines = report.split(/\r?\n/)
|
|
228
|
+
const sectionIndexes = new Map(
|
|
229
|
+
lines.map((line, index) => [line.trim(), index]).filter(([line]) => REFLECTION_SECTIONS.includes(line)),
|
|
230
|
+
)
|
|
231
|
+
if (!REFLECTION_SECTIONS.every((section) => sectionIndexes.has(section))) return false
|
|
232
|
+
const decisionStart = sectionIndexes.get("## Decision") + 1
|
|
233
|
+
const nextHeading = lines.findIndex((line, index) => index >= decisionStart && /^##\s+/.test(line.trim()))
|
|
234
|
+
const decisionLines = lines.slice(decisionStart, nextHeading === -1 ? lines.length : nextHeading)
|
|
235
|
+
const firstDecisionLine = decisionLines.find((line) => line.trim().length > 0)
|
|
236
|
+
return !!firstDecisionLine && /^(?:RESUME|ESCALATE)\s*:/i.test(firstDecisionLine.trim()) &&
|
|
237
|
+
report.includes(`Reflection token: ${breaker.reflectionToken}`)
|
|
238
|
+
} catch {
|
|
239
|
+
return false
|
|
240
|
+
}
|
|
241
|
+
}
|
|
242
|
+
|
|
243
|
+
function completeReflection(breakerFile, reportFile) {
|
|
244
|
+
const breaker = loadBreakerState(breakerFile)
|
|
245
|
+
if (!isValidReflection(reportFile, breaker)) return false
|
|
246
|
+
const completedAt = Math.max(Date.now(), (breaker.reflectionRequestedAt || 0) + 1)
|
|
247
|
+
breaker.lastReflection = completedAt
|
|
248
|
+
breaker.lastReflectionSignature = breaker.reflectionSignature
|
|
249
|
+
breaker.reflectionPending = false
|
|
250
|
+
breaker.reflectionRequestedAt = null
|
|
251
|
+
breaker.reflectionToken = null
|
|
252
|
+
breaker.reflectionSignature = null
|
|
253
|
+
saveJSON(breakerFile, breaker)
|
|
254
|
+
return true
|
|
255
|
+
}
|
|
256
|
+
|
|
257
|
+
function defaultCompliance() {
|
|
258
|
+
return {
|
|
259
|
+
sessionStart: Date.now(),
|
|
260
|
+
totalEdits: 0,
|
|
261
|
+
verifiedEdits: 0,
|
|
262
|
+
circuitBreakerTrips: 0,
|
|
263
|
+
reflectionsForced: 0,
|
|
264
|
+
}
|
|
265
|
+
}
|
|
266
|
+
|
|
267
|
+
function extractFailureSignature(text) {
|
|
268
|
+
return text.slice(0, 100).replace(/\s+/g, " ").trim()
|
|
269
|
+
}
|
|
270
|
+
|
|
271
|
+
function getAvailableScripts(cwd) {
|
|
272
|
+
const pkgPath = join(cwd, "package.json")
|
|
273
|
+
if (!existsSync(pkgPath)) return {}
|
|
274
|
+
try {
|
|
275
|
+
return JSON.parse(readFileSync(pkgPath, "utf8")).scripts || {}
|
|
276
|
+
} catch {
|
|
277
|
+
return {}
|
|
278
|
+
}
|
|
279
|
+
}
|
|
280
|
+
|
|
281
|
+
function runVerification(cwd) {
|
|
282
|
+
const scripts = getAvailableScripts(cwd)
|
|
283
|
+
const commands = [
|
|
284
|
+
{ cmd: "npm test", script: "test" },
|
|
285
|
+
{ cmd: "npm run lint", script: "lint" },
|
|
286
|
+
{ cmd: "npm run build", script: "build" },
|
|
287
|
+
].filter((c) => scripts[c.script])
|
|
288
|
+
|
|
289
|
+
if (commands.length === 0) {
|
|
290
|
+
return { allPassed: true, skipped: true, results: [] }
|
|
291
|
+
}
|
|
292
|
+
|
|
293
|
+
const results = []
|
|
294
|
+
for (const { cmd } of commands) {
|
|
295
|
+
try {
|
|
296
|
+
execSync(cmd, { cwd, stdio: "pipe", timeout: 60000 })
|
|
297
|
+
results.push({ command: cmd, success: true })
|
|
298
|
+
} catch (e) {
|
|
299
|
+
results.push({ command: cmd, success: false, error: (e.stdout || e.message || "").toString().slice(0, 200) })
|
|
300
|
+
}
|
|
301
|
+
}
|
|
302
|
+
return { allPassed: results.every((r) => r.success), skipped: false, results }
|
|
303
|
+
}
|
|
304
|
+
|
|
305
|
+
/**
|
|
306
|
+
* Records a verification failure against the circuit breaker and returns the
|
|
307
|
+
* action to take: allow (with retries remaining), force_reflection (3rd
|
|
308
|
+
* failure on this signature), or hard_lock (same signature repeats after a
|
|
309
|
+
* reflection was recorded).
|
|
310
|
+
*/
|
|
311
|
+
function tripBreaker(breakerFile, signature) {
|
|
312
|
+
const breaker = loadBreakerState(breakerFile)
|
|
313
|
+
|
|
314
|
+
if (!breaker.failures[signature]) {
|
|
315
|
+
breaker.failures[signature] = { count: 0, firstSeen: Date.now() }
|
|
316
|
+
}
|
|
317
|
+
const entry = breaker.failures[signature]
|
|
318
|
+
entry.count++
|
|
319
|
+
entry.lastSeen = Date.now()
|
|
320
|
+
|
|
321
|
+
if (entry.count >= 3) {
|
|
322
|
+
if (
|
|
323
|
+
breaker.lastReflection &&
|
|
324
|
+
breaker.lastReflection > entry.firstSeen &&
|
|
325
|
+
breaker.lastReflectionSignature === signature
|
|
326
|
+
) {
|
|
327
|
+
breaker.hardLock = true
|
|
328
|
+
saveJSON(breakerFile, breaker)
|
|
329
|
+
return { action: "hard_lock", count: entry.count }
|
|
330
|
+
}
|
|
331
|
+
if (!breaker.reflectionPending) {
|
|
332
|
+
breaker.reflectionPending = true
|
|
333
|
+
breaker.reflectionRequestedAt = Date.now()
|
|
334
|
+
breaker.reflectionToken = randomUUID().replace(/-/g, "")
|
|
335
|
+
breaker.reflectionSignature = signature
|
|
336
|
+
}
|
|
337
|
+
saveJSON(breakerFile, breaker)
|
|
338
|
+
return { action: "force_reflection", count: entry.count, reflectionToken: breaker.reflectionToken }
|
|
339
|
+
}
|
|
340
|
+
|
|
341
|
+
saveJSON(breakerFile, breaker)
|
|
342
|
+
return { action: "allow", count: entry.count, remaining: 3 - entry.count }
|
|
343
|
+
}
|
|
344
|
+
|
|
345
|
+
function recordCompliance(complianceFile, mutate) {
|
|
346
|
+
const compliance = loadJSON(complianceFile, defaultCompliance())
|
|
347
|
+
mutate(compliance)
|
|
348
|
+
saveJSON(complianceFile, compliance)
|
|
349
|
+
}
|
|
350
|
+
|
|
351
|
+
export const HarnessEnforcement = async ({ client, directory }) => {
|
|
352
|
+
// State is resolved per real workspace and per opencode session. The CJS
|
|
353
|
+
// hooks use the same workspaces/<key>/state/sessions/<id> contract, while
|
|
354
|
+
// the old flat plugin state is intentionally left untouched because it is
|
|
355
|
+
// impossible to attribute safely to one workspace or session.
|
|
356
|
+
const workspace = resolve(directory || process.cwd())
|
|
357
|
+
migrateLegacyFlatState(getStateDir(workspace))
|
|
358
|
+
const activeVerifications = new Set()
|
|
359
|
+
|
|
360
|
+
function pathsFor(sessionID) {
|
|
361
|
+
const stateDir = getSessionDir(workspace, sessionID)
|
|
362
|
+
return {
|
|
363
|
+
stateDir,
|
|
364
|
+
editStateFile: join(stateDir, "edit-state.json"),
|
|
365
|
+
breakerFile: join(stateDir, "circuit-breaker.json"),
|
|
366
|
+
complianceFile: join(stateDir, "compliance.json"),
|
|
367
|
+
reflectionFile: reflectionReportPath(stateDir),
|
|
368
|
+
}
|
|
369
|
+
}
|
|
370
|
+
|
|
371
|
+
function resetSession(sessionID) {
|
|
372
|
+
const { stateDir } = pathsFor(sessionID)
|
|
373
|
+
const sessionsRoot = resolve(join(getStateRoot(workspace), "sessions"))
|
|
374
|
+
if (!isWithin(sessionsRoot, stateDir) || resolve(stateDir) === sessionsRoot) return
|
|
375
|
+
try { rmSync(stateDir, { recursive: true, force: true }) } catch { /* fail open */ }
|
|
376
|
+
}
|
|
377
|
+
|
|
378
|
+
function toolTarget(input, output) {
|
|
379
|
+
const args = { ...((output && output.args) || {}), ...((input && input.args) || {}) }
|
|
380
|
+
return args.filePath || args.file_path || args.path || args.filename || args.file || null
|
|
381
|
+
}
|
|
382
|
+
|
|
383
|
+
function patchTargets(input, output) {
|
|
384
|
+
const args = { ...((output && output.args) || {}), ...((input && input.args) || {}) }
|
|
385
|
+
if (typeof args.patchText !== "string") return []
|
|
386
|
+
return [...args.patchText.matchAll(/^\*\*\*\s+(?:(?:Add|Update|Delete) File|Move to):[ \t]*([^\r\n]+)$/gm)]
|
|
387
|
+
.map((match) => match[1].trim())
|
|
388
|
+
}
|
|
389
|
+
|
|
390
|
+
function isReflectionWrite(input, output, reportFile) {
|
|
391
|
+
if (!EDIT_TOOLS.has(input.tool)) return false
|
|
392
|
+
if (input.tool === "apply_patch") {
|
|
393
|
+
const targets = patchTargets(input, output)
|
|
394
|
+
return targets.length > 0 && targets.every((patchTarget) => samePath(resolve(workspace, patchTarget), reportFile))
|
|
395
|
+
}
|
|
396
|
+
const target = toolTarget(input, output)
|
|
397
|
+
return Boolean(target && samePath(resolve(workspace, target), reportFile))
|
|
398
|
+
}
|
|
399
|
+
|
|
400
|
+
return {
|
|
401
|
+
"tool.execute.before": async (input, output) => {
|
|
402
|
+
if (!EDIT_TOOLS.has(input.tool)) return
|
|
403
|
+
const { breakerFile, reflectionFile } = pathsFor(input.sessionID)
|
|
404
|
+
const breaker = loadBreakerState(breakerFile)
|
|
405
|
+
if (breaker.hardLock) {
|
|
406
|
+
throw new Error(
|
|
407
|
+
`Harness circuit breaker hard-locked after a repeat failure post-reflection. ` +
|
|
408
|
+
`Delete "${breakerFile}" or start a new session to reset.`,
|
|
409
|
+
)
|
|
410
|
+
}
|
|
411
|
+
if (breaker.reflectionPending && !isReflectionWrite(input, output, reflectionFile)) {
|
|
412
|
+
throw new Error(
|
|
413
|
+
`Harness reflection is required before another code edit. ` +
|
|
414
|
+
`Write a valid reflection report to "${reflectionFile}" first.`,
|
|
415
|
+
)
|
|
416
|
+
}
|
|
417
|
+
},
|
|
418
|
+
|
|
419
|
+
"tool.execute.after": async (input, output) => {
|
|
420
|
+
if (!EDIT_TOOLS.has(input.tool)) return
|
|
421
|
+
const { editStateFile, breakerFile, complianceFile, reflectionFile } = pathsFor(input.sessionID)
|
|
422
|
+
|
|
423
|
+
// The reflection artifact is a protocol write, not a code edit. It is
|
|
424
|
+
// accepted only when it contains the current token and all required
|
|
425
|
+
// sections; this is the only path that records lastReflection.
|
|
426
|
+
if (isReflectionWrite(input, output, reflectionFile)) {
|
|
427
|
+
completeReflection(breakerFile, reflectionFile)
|
|
428
|
+
return
|
|
429
|
+
}
|
|
430
|
+
|
|
431
|
+
const state = normalizeEditState(loadJSON(editStateFile, defaultEditState()))
|
|
432
|
+
state.lastEditTime = Date.now()
|
|
433
|
+
state.verificationPending = true
|
|
434
|
+
state.followUpPending = false
|
|
435
|
+
state.followUpDeliveryPending = false
|
|
436
|
+
state.followUpMessage = null
|
|
437
|
+
state.lastFollowUpError = null
|
|
438
|
+
state.editsSinceVerification++
|
|
439
|
+
saveJSON(editStateFile, state)
|
|
440
|
+
recordCompliance(complianceFile, (c) => c.totalEdits++)
|
|
441
|
+
},
|
|
442
|
+
|
|
443
|
+
event: async ({ event }) => {
|
|
444
|
+
const properties = event && event.properties
|
|
445
|
+
const sessionID = properties && (properties.sessionID || (properties.info && properties.info.id))
|
|
446
|
+
if (event.type === "session.created") {
|
|
447
|
+
resetSession(sessionID)
|
|
448
|
+
return
|
|
449
|
+
}
|
|
450
|
+
if (event.type === "session.deleted") {
|
|
451
|
+
resetSession(sessionID)
|
|
452
|
+
return
|
|
453
|
+
}
|
|
454
|
+
if (event.type !== "session.idle") return
|
|
455
|
+
if (activeVerifications.has(sessionID || "default")) return
|
|
456
|
+
|
|
457
|
+
const activeSession = sessionID || "default"
|
|
458
|
+
activeVerifications.add(activeSession)
|
|
459
|
+
try {
|
|
460
|
+
const { editStateFile, breakerFile, complianceFile, reflectionFile } = pathsFor(activeSession)
|
|
461
|
+
// This fallback makes the lifecycle complete even if the host omits
|
|
462
|
+
// tool input args on the artifact's after hook.
|
|
463
|
+
completeReflection(breakerFile, reflectionFile)
|
|
464
|
+
|
|
465
|
+
const state = normalizeEditState(loadJSON(editStateFile, defaultEditState()))
|
|
466
|
+
if (state.followUpPending) return
|
|
467
|
+
|
|
468
|
+
if (state.followUpDeliveryPending && state.followUpMessage) {
|
|
469
|
+
try {
|
|
470
|
+
await client.session.prompt({
|
|
471
|
+
path: { id: activeSession },
|
|
472
|
+
body: { parts: [{ type: "text", text: state.followUpMessage }] },
|
|
473
|
+
})
|
|
474
|
+
state.followUpPending = true
|
|
475
|
+
state.followUpDeliveryPending = false
|
|
476
|
+
state.followUpMessage = null
|
|
477
|
+
state.lastFollowUpError = null
|
|
478
|
+
saveJSON(editStateFile, state)
|
|
479
|
+
} catch (error) {
|
|
480
|
+
state.lastFollowUpError = String(error && (error.message || error)).slice(0, 300)
|
|
481
|
+
saveJSON(editStateFile, state)
|
|
482
|
+
}
|
|
483
|
+
return
|
|
484
|
+
}
|
|
485
|
+
|
|
486
|
+
if (!state.verificationPending || state.editsSinceVerification === 0) return
|
|
487
|
+
|
|
488
|
+
const { allPassed, skipped, results } = runVerification(workspace)
|
|
489
|
+
|
|
490
|
+
if (skipped || allPassed) {
|
|
491
|
+
state.verificationPending = false
|
|
492
|
+
state.followUpPending = false
|
|
493
|
+
state.editsSinceVerification = 0
|
|
494
|
+
saveJSON(editStateFile, state)
|
|
495
|
+
recordCompliance(complianceFile, (c) => c.verifiedEdits++)
|
|
496
|
+
return
|
|
497
|
+
}
|
|
498
|
+
|
|
499
|
+
const failing = results.find((r) => !r.success)
|
|
500
|
+
const signature = extractFailureSignature(`${failing.command}: ${failing.error || ""}`)
|
|
501
|
+
const trip = tripBreaker(breakerFile, signature)
|
|
502
|
+
state.followUpPending = true
|
|
503
|
+
saveJSON(editStateFile, state)
|
|
504
|
+
|
|
505
|
+
let text
|
|
506
|
+
if (trip.action === "hard_lock") {
|
|
507
|
+
text =
|
|
508
|
+
`Harness: same verification failure ("${failing.command}") returned after a reflection was ` +
|
|
509
|
+
`already recorded. The circuit breaker is now hard-locked - edits are blocked until it is reset.`
|
|
510
|
+
recordCompliance(complianceFile, (c) => c.circuitBreakerTrips++)
|
|
511
|
+
} else if (trip.action === "force_reflection") {
|
|
512
|
+
const { stateDir } = pathsFor(activeSession)
|
|
513
|
+
text =
|
|
514
|
+
`Harness: "${failing.command}" has now failed 3 times with the same error. Stop and reflect ` +
|
|
515
|
+
`before retrying. Write a report to "${reflectionReportPath(stateDir)}" with sections ` +
|
|
516
|
+
`## Goal, ## Failed Attempts, ## Verified Facts, ## Diagnosis, and ## Decision. ` +
|
|
517
|
+
`The Decision must begin with RESUME: or ESCALATE:. Include reflection token: ${trip.reflectionToken}. ` +
|
|
518
|
+
`After the report is recorded, resume with a different approach. Failure: ${failing.error || "(no output captured)"}`
|
|
519
|
+
recordCompliance(complianceFile, (c) => {
|
|
520
|
+
c.circuitBreakerTrips++
|
|
521
|
+
c.reflectionsForced++
|
|
522
|
+
})
|
|
523
|
+
} else {
|
|
524
|
+
text =
|
|
525
|
+
`Harness: verification failed (${trip.remaining} retries before a forced reflection). ` +
|
|
526
|
+
`"${failing.command}" - ${failing.error || "(no output captured)"}. Fix it and this will re-run automatically.`
|
|
527
|
+
}
|
|
528
|
+
|
|
529
|
+
try {
|
|
530
|
+
await client.session.prompt({ path: { id: activeSession }, body: { parts: [{ type: "text", text }] } })
|
|
531
|
+
} catch (error) {
|
|
532
|
+
// A failed delivery is retryable on the next idle event, but the
|
|
533
|
+
// saved message means retrying does not run verification again or
|
|
534
|
+
// increment the circuit-breaker count.
|
|
535
|
+
const failedState = normalizeEditState(loadJSON(editStateFile, defaultEditState()))
|
|
536
|
+
failedState.followUpPending = false
|
|
537
|
+
failedState.followUpDeliveryPending = true
|
|
538
|
+
failedState.followUpMessage = text
|
|
539
|
+
failedState.lastFollowUpError = String(error && (error.message || error)).slice(0, 300)
|
|
540
|
+
saveJSON(editStateFile, failedState)
|
|
541
|
+
}
|
|
542
|
+
} finally {
|
|
543
|
+
activeVerifications.delete(activeSession)
|
|
544
|
+
}
|
|
545
|
+
},
|
|
546
|
+
}
|
|
547
|
+
}
|
|
@@ -1,17 +1,9 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "harness-enforcement",
|
|
3
|
-
"description": "
|
|
4
|
-
"version": "0.
|
|
5
|
-
"
|
|
6
|
-
"license": "Apache-2.0",
|
|
3
|
+
"description": "Repository inventory for the opencode enforcement plugin and its standalone helper",
|
|
4
|
+
"version": "0.4.0",
|
|
5
|
+
"runtime": "index.mjs",
|
|
7
6
|
"hooks": {
|
|
8
|
-
"
|
|
9
|
-
"preComplete": "hooks/pre-complete.js"
|
|
10
|
-
},
|
|
11
|
-
"config": {
|
|
12
|
-
"verificationRequired": true,
|
|
13
|
-
"circuitBreakerEnabled": true,
|
|
14
|
-
"maxRetries": 3,
|
|
15
|
-
"verificationCommands": ["npm test", "npm run lint", "npm run build"]
|
|
7
|
+
"verify": "hooks/verify.js"
|
|
16
8
|
}
|
|
17
9
|
}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "harness-everything",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.4.0",
|
|
4
4
|
"description": "Harness OS: An Orchestrated Agent Operating System for AI-driven Development",
|
|
5
5
|
"main": "harness-everything/scripts/bootstrap.js",
|
|
6
6
|
"bin": {
|
|
@@ -10,13 +10,23 @@
|
|
|
10
10
|
"test": "node self-evolve/scripts/self-regression.js",
|
|
11
11
|
"self-regression": "node self-evolve/scripts/self-regression.js",
|
|
12
12
|
"test:mechanism": "node ci/mechanism-test.js",
|
|
13
|
-
"test:
|
|
13
|
+
"test:mutations": "node ci/mutation-check.js",
|
|
14
|
+
"test:tdd-quality": "node ci/mechanism-2s-tdd-quality.test.js",
|
|
15
|
+
"tdd:quality": "node tdd/scripts/quality-gate.js",
|
|
16
|
+
"test:consistency": "node ci/consistency-check.js && node ci/doc-capability-consistency.test.js",
|
|
17
|
+
"test:docs:capabilities": "node ci/doc-capability-consistency.test.js",
|
|
14
18
|
"test:references": "node ci/reference-check.js",
|
|
15
|
-
"test:release": "node ci/release-consistency-check.js
|
|
19
|
+
"test:release": "node ci/release-consistency-check.js && node ci/mechanism-2x-release-automation.test.js && node ci/mechanism-2x-release-commit-message.test.js",
|
|
16
20
|
"test:install": "node scripts/verify-install.js",
|
|
17
21
|
"test:collision": "node ci/description-collision.js",
|
|
18
22
|
"test:routing:skills": "node ci/skill-routing-check.js",
|
|
19
23
|
"test:routing": "node ci/routing-check.js",
|
|
24
|
+
"test:routing:invariants": "node ci/invariant-routing.test.js && node ci/router-workflow-plan.test.js",
|
|
25
|
+
"plugin:sync": "node scripts/sync-openai-plugin.js",
|
|
26
|
+
"test:plugin:openai": "node ci/openai-plugin-package.test.js && node ci/openai-plugin-runtime.test.js",
|
|
27
|
+
"plugin:submission:build": "node scripts/build-openai-submission.js",
|
|
28
|
+
"test:plugin:submission": "node ci/openai-plugin-submission.test.js",
|
|
29
|
+
"test:platform:compatibility": "node ci/mechanism-2w-platform-compatibility.test.js",
|
|
20
30
|
"benchmark:scaffold": "node benchmarks/run.js scaffold",
|
|
21
31
|
"benchmark:status": "node benchmarks/run.js status",
|
|
22
32
|
"eval:behavioral": "node behavioral-evals/run.js run",
|
|
@@ -41,6 +51,9 @@
|
|
|
41
51
|
"bugs": {
|
|
42
52
|
"url": "https://github.com/dyphn1/Harness-everything/issues"
|
|
43
53
|
},
|
|
54
|
+
"publishConfig": {
|
|
55
|
+
"access": "public"
|
|
56
|
+
},
|
|
44
57
|
"devDependencies": {
|
|
45
58
|
"js-yaml": "^5.4.1",
|
|
46
59
|
"@commitlint/cli": "^19.0.0",
|
|
@@ -0,0 +1,39 @@
|
|
|
1
|
+
{
|
|
2
|
+
"name": "harness-everything",
|
|
3
|
+
"version": "0.4.0",
|
|
4
|
+
"description": "Invariant-first engineering harness for routing, verification, and failure recovery while preserving agent autonomy.",
|
|
5
|
+
"author": {
|
|
6
|
+
"name": "Miya Daniel",
|
|
7
|
+
"url": "https://github.com/dyphn1"
|
|
8
|
+
},
|
|
9
|
+
"homepage": "https://github.com/dyphn1/Harness-everything",
|
|
10
|
+
"repository": "https://github.com/dyphn1/Harness-everything",
|
|
11
|
+
"license": "Apache-2.0",
|
|
12
|
+
"keywords": [
|
|
13
|
+
"agent-harness",
|
|
14
|
+
"orchestration",
|
|
15
|
+
"tdd",
|
|
16
|
+
"verification"
|
|
17
|
+
],
|
|
18
|
+
"skills": "./skills/",
|
|
19
|
+
"hooks": "./hooks/hooks.json",
|
|
20
|
+
"interface": {
|
|
21
|
+
"displayName": "Harness Everything",
|
|
22
|
+
"shortDescription": "Route, verify, and recover coding work",
|
|
23
|
+
"longDescription": "Harness Everything adds reusable software-engineering workflows for task routing, test-driven development, verification, security review, documentation, failure recovery, and related coding-agent work. It keeps a small set of cross-cutting invariants while allowing the agent to choose, combine, reorder, or skip domain skills as appropriate.",
|
|
24
|
+
"developerName": "Miya Daniel",
|
|
25
|
+
"category": "Developer Tools",
|
|
26
|
+
"capabilities": [
|
|
27
|
+
"Interactive",
|
|
28
|
+
"Read",
|
|
29
|
+
"Write"
|
|
30
|
+
],
|
|
31
|
+
"websiteURL": "https://github.com/dyphn1/Harness-everything",
|
|
32
|
+
"privacyPolicyURL": "https://github.com/dyphn1/Harness-everything/blob/main/PRIVACY.md",
|
|
33
|
+
"termsOfServiceURL": "https://github.com/dyphn1/Harness-everything/blob/main/TERMS.md",
|
|
34
|
+
"defaultPrompt": [
|
|
35
|
+
"Route this software task with Harness and use only the skills that help.",
|
|
36
|
+
"Review this change and verify it with objective evidence before claiming completion."
|
|
37
|
+
]
|
|
38
|
+
}
|
|
39
|
+
}
|