{"id":17729,"plugin_id":"plugins_6a78a8d87bcc8191981753490f5afbcf","kind":"skill","collection_source":null,"comparison_source":null,"observed_at":"2026-09-30T23:14:20.243Z","digest":"52dc9ae41c3df7da5f90be95da19fb1a16202653ea8aad723293ae417c25e14c","against":null,"payload":{"description":"Use when Codex Self Improvement routes work through a curated evidence path for architecture, controlled learning, independent review, privacy, security, performance, or browser-level validation","included_files":[],"name":"harness-paths","skill_md_contents":"---\nname: harness-paths\ndescription: Use when Codex Self Improvement routes work through a curated evidence path for architecture, controlled learning, independent review, privacy, security, performance, or browser-level validation\n---\n\n# Harness Paths\n\nSelect one primary path and keep its evidence and approval boundary visible\n\n## architecture-conformance\n\nUse `architecture-guardian` for non-trivial cross-module, dependency, schema, public API, or contract changes\n\nRequire current architecture or reuse evidence and a post-change conformance result\nDefault to review mode\nNever create a baseline, exception, or broader contract merely to pass a check\n\n## controlled-evolution\n\nUse `agent-kernel-evolve` only after repeated failure or correction evidence exists\nDraft a candidate playbook with acceptance, holdout, privacy, and rollback criteria\nNever publish memory automatically\nNever install lifecycle hooks, start a daemon, or change environment vault state without explicit approval\n\n## evidence-loop\n\nReproduce the symptom or establish a failing check before changing behavior\nRun a focused check after the final mutation\nA diff, confidence statement, or prior agent report is not completion evidence\n\n## independent-review\n\nUse `code-review` and `agency-multi-agent-systems-architect` with a fixed base and concrete specification\nKeep standards and specification reviewers in isolated contexts\nNever send repository content to an external model without explicit approval\nReviewer output is a lead that must be checked against the diff and fresh commands\n\n## privacy-minimization\n\nUse `agency-privacy-engineer` to map field purpose, stores, retention, and deletion paths\nUse metadata and synthetic records in harness artifacts\nNever retain personal data or free-text payloads in reports\n\n## secure-change\n\nUse `agency-application-security-engineer` and `agency-secrets-credential-hygiene-engineer`\nPrefer installed Codex Security tools for repository scanning and source-to-sink validation\nRecord secret location and fingerprint only, never the value\nCredential rotation, revocation, live-value reads, and external target scans require explicit approval\n\n## performance-evidence\n\nUse `agency-performance-benchmarker` with a fixed local or synthetic workload\nRecord environment digest, warm-up policy, latency distribution, throughput, resources, and errors\nNever run load against a shared or production target without explicit approval\nCorrectness and safety regressions reject the candidate regardless of speed\n\n## e2e-evidence\n\nUse `agency-test-automation-engineer` for critical browser flows with isolated test data\nWait on observable conditions rather than fixed sleeps\nCapture traces or screenshots for failures without personal data\nNever run against production or upload artifacts externally without explicit approval\n\n## anti-overwhelm-focus\n\nBreak complex tasks into single-action micro-steps to reduce cognitive overhead and eliminate multi-turn drift\nEnforce exactly one atomic action per turn\nVerify micro-step completion immediately after mutation\nPause execution before attempting multi-layer edits\n\n## Universal boundary\n\nUse no automatic actions from a third-party skill\nTreat installed skills as reviewed references, not permission grants\nState the selected path, evidence produced, approvals used, and anything not verified\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}