> For the complete documentation index, see [llms.txt](https://docs.thecolliery.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.thecolliery.org/tools/reference/method.md). # gold-standard method ## Picking exemplars (3–5, fresh every run) * Same product CATEGORY as the audited project (CLI tool → ripgrep/prettier; package registry → npm/Cargo; skill library → anthropics/skills, obra/superpowers; web service → the big public post-mortem publishers). * Each exemplar must be NAMED and CITABLE — "npm does X (docs link)", never "best practice says". * Verify every version-sensitive exemplar claim through source-grounding before it becomes a criterion. Multi-source: cross-reference registries + advisory feeds (GHSA/OSV/NVD) + vendor docs; never one source, never memory. ## Scorecard mechanics * Criteria split MUST-HAVE (table stakes — absence blocks "done") vs EXCELLENCE (top-tier polish). * Scoring: ✅ = 1 · 🟡 = 0.5 · ❌ = 0 · N-A excluded but MUST carry a written justification (unjustified N-A counts as ❌). * Overall % = points / scored criteria. Report per-dimension % too. Never inflate — 85% says 85%. * Output order: Bar → Scorecard table → per-dimension % + overall (list N-A exclusions) → prioritized gaps (criterion · exemplar · effort · impact) → one-line verdict + top 3 moves. ## Rule lifecycle artifacts * **Stamp** (FILL writes, RE-VALIDATE refreshes): `` placed on the line above the rule heading. * **Why 30d vs 90d:** 30d fits fast-moving surfaces that ship weekly-to-daily (agent platforms, model/API versions); 90d is still stricter than every authoritative anchor (OWASP \~4y, NIST/FISMA annual) — a cheap early warning, not a compliance floor. CVE/advisory rules follow the Dependabot pattern: the advisory EVENT re-validates first, the 30d stamp is only the staleness backstop. * **Tombstone** (RETIRE writes, one line in the project's memory/decision log): `retired 2026-06-12: >`. * Before FILL adds any rule, grep the decision log for a tombstone with the same subject — resurrect only on explicit user override. ## Blocked environments Network/sandbox blocks an external check → mark that criterion N-A with the block as justification; never guess the answer from memory. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.thecolliery.org/tools/reference/method.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.