reference-doc-routing
Skill0 starsChanged 4 months agoMITView on GitHub
Use when a task might need engineering-design, architecture, testing, access-control, design-system, API, release, data-model, subagent, or worktree reference documents. Helps pick the ones that matter, judge how much authority each has, read only the relevant sections, and pass concise labeled context to subagents. Packages the reference catalog, the engineering-design decision aid, and the repository documentation templates.
1 other project uses a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- ArcanEdge-AI/coding-agent-playbook-codexedited, first seen Aug 17, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 1
- ArcanEdge-AI/coding-agent-playbook-codex★ 313% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository28 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- aws-local-preflightSkill
- feature-branch-lifecycleSkill
- handoffSkill
- legacy-path-retirementSkill
- local-ci-preflightSkill
- multi-role-repo-reviewSkill
- multi-session-coordinationSkill
- platform-documentationSkill
- production-api-integrationSkill
- production-architecture-reviewSkill
- production-code-qualitySkill
- production-database-integritySkill
- production-devops-releaseSkill
- production-documentation-onboardingSkill
- production-pass-orchestratorSkill
- production-performance-reviewSkill
- production-readiness-reportSkill
- production-reliability-observabilitySkill
- production-security-reviewSkill
- production-testing-qaSkill
- production-ux-accessibilitySkill
- senior-code-reviewSkill
- session-cleanupSkill
- subagent-orchestrationSkill
- task-graph-orchestrationSkill
- technical-debt-auditorSkill
- worktree-lifecycleSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Verify anything implementation-relevant against the code: before you rely on it.
- Distinctive: line 42
Ask for evidence, not agreement: the subagent should return the paths and symbols it checked, not a restatement of what you sent.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reference Doc Routing
A reference document tells you what someone intended. The code tells you what is true. Route documents so you get the intent without inheriting the drift.
Full detail: references/reference-doc-routing.md. This package also carries references/README.md, the catalog of every reference and template the playbook ships by owning skill; references/engineering-design.md, the decision questions for non-trivial design choices; and references/templates/, starter files for repository documentation. A reference owned by another skill is reached by using that skill, not by copying its file.
Workflow
- Identify what actually needs reference context. Much of a task is answerable from the code alone. Reading a document you did not need costs context the code needed.
- Find the candidates — global references under the Claude Code home, and repository docs.
- Classify each one:
- Authoritative — instructions make it the source of truth. Follow it; raise conflicts rather than deviating.
- Advisory — useful, but current code, tests, or user instructions override it.
- Historical — may describe decisions or behavior that no longer exist.
- Read narrowly.
Grepfor the symbol, endpoint, table, or concept, then read around the hits. Read end to end only when the task really is about the whole subject. - Check for drift. Does it reference files, symbols, or endpoints that no longer exist? Do its examples use an API shape the code has moved past? Does it predate a migration visible in the history? Does it contradict the tests?
- Verify anything implementation-relevant against the code before you rely on it.
- Report conflicts between a document and primary evidence. Do not silently pick one — work out whether the doc is stale, the code is a bug, or the code is deliberately working around something. Authority and evidence answer different questions: the instruction hierarchy and authorized requirements govern what the behavior must be; code, tests, schemas, and runtime observations establish what it is. A passing test does not override an authorized requirement. Update a stale document only within the task's authority; otherwise flag it.
Passing documents to a subagent
A subagent has a fresh context and sees only what you send. Send the path, the section, and the label:
Reference documents:
- docs/api-contracts.md, "Checkout" section — authoritative. Follow these
request and response shapes exactly.
- docs/architecture.md, "Payments" section — advisory, last updated before the
Stripe migration. Verify anything you take from it against current code.
Verify implementation-relevant claims against the current code. Do not
summarize unrelated sections.An unlabeled document is treated as truth. That is how you get back confident work built on a stale premise.
Ask for evidence, not agreement: the subagent should return the paths and symbols it checked, not a restatement of what you sent.
Primary evidence
Current code, tests, schemas, configuration, logs, build output, typecheck output, runtime behavior, and authoritative external documentation.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

