librarian
Skill2 starsChanged todayMITView on GitHub
Creates and maintains documentation, READMEs, ADRs, and API references. Use when documentation is missing, outdated, or after code changes that affect documented behavior. The Librarian ensures that knowledge outlives the developer who created it.
- Reads credentials
- Installs packages
21 other projects use a version of this.
Adopted by
Most recently taken up by
- JetBrains/thinkrailtheir own version, first seen Nov 15, 2026
- jwm-axoni/claude-workbenchtheir own version, first seen Nov 8, 2026
- therealtinhtute/mono-harnesstheir own version, first seen Nov 1, 2026
- skills-agents-co/skills-and-agents-librarytheir own version, first seen Oct 15, 2026
- ivanrvpereira/.agentstheir own version, first seen Oct 13, 2026
- mitsuhiko/agent-commandstheir own version, first seen Oct 8, 2026
Used with
Projects with it also use bro, retro and 1 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
MicrosoftSkillbro
Re-explain the previous assistant message in a much simpler way, for when the reply made you go 'bro what'. Use /bro to get a simpler version of the last answer in Plain Language + Chicago Manual of Style, with diagrams where they help.
Also in 3 projects that use this
AWSSkillretro
Conduct a retrospective on a coding session.
Also in 3 projects that use this
GoogleSkilltdd
Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.
Also in 4 projects that use this
Other versions21 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 21
- JetBrains/thinkrail★ 8.3ksame name
- mitsuhiko/agent-commands★ 3.2ksame name
- mitsuhiko/agent-stuff★ 3.2ksame name
- diodeinc/pcb★ 477same name
- ngmeyer/librarian-mcp★ 29same name
- roerohan/skills★ 29same name
- ivanrvpereira/.agents★ 19same name
- therealtinhtute/mono-harness★ 5same name
- SVerITG/Metis★ 4same name
- jwm-axoni/claude-workbench★ 2same name
11 more not listed
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 repository60 other files it gives its agents.
- AGENTS.mdAGENTS.md
- accessibility-auditorSkill
- architectSkill
- auditorSkill
- awsSkill
- bug-fix-teammateSkill
- builderSkill
- cleanup-specialistSkill
- code-reviewSkill
- commit-messagesSkill
- concept-explainerSkill
- context-economySkill
- cross-provider-translationSkill
- datadogSkill
- debuggerSkill
- debugging-and-error-recoverySkill
- debugging-tutorSkill
- dockerSkill
- doormanSkill
- elasticsearchSkill
- emailSkill
- envoySkill
- githubSkill
- gitlabSkill
- goal-refinementSkill
- heraldSkill
- implementation-plannerSkill
- inspectorSkill
- institutional-knowledgeSkill
- issue-managerSkill
- jiraSkill
- kubernetesSkill
- linearSkill
- mailmanSkill
- mediatorSkill
- mongodbSkill
- mysqlSkill
- onboarding-planSkill
- operatorSkill
- oracleSkill
- postgresSkill
- pull-request-assistantSkill
- redisSkill
- release-notesSkill
- rememberSkill
- reviewerSkill
- runtime-healthSkill
- scribeSkill
- security-and-hardeningSkill
- sentinelSkill
- sentrySkill
- society-integritySkill
- spec-driven-developmentSkill
- stewardSkill
- strategistSkill
- telegramSkill
- test-driven-developmentSkill
- testerSkill
- validation-and-enforcementSkill
- wardenSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. 0 are common, 14 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 21 edited versions we read keep these.
14
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
When a module or feature has no documentation
- Distinctive: line 17
After code changes that affect a documented API or workflow
- Distinctive: line 18
When a new team member would need more than 30 minutes to understand a component
- Distinctive: line 19
After a significant architectural decision (produce an ADR)
- Distinctive: line 20
When onboarding a new contributor
- Very specific: line 70
Use GitHub Flavored Markdown with relative links (docs/CONTRIBUTING.md), never absolute URLs for in-repo files — links must work when the …
- Distinctive: line 71
Use proper heading structure so GitHub's auto-generated table of contents works
- Distinctive: line 72
Keep content under 500 KiB — GitHub truncates beyond that
- Distinctive: line 73
Add badges for build status, version, license if appropriate
- Distinctive: line 186
After code changes, identify stale documentation:
- Distinctive: line 188
Read the changed files
- Distinctive: line 189
Search for documentation that references those files, functions, or behaviors
- Distinctive: line 191
Update it to match the new behavior
- Distinctive: line 199
When a postmortem is finalized:
- Distinctive: line 203
Update relevant documentation (READMEs, runbooks, ADRs) to reflect lessons learned
“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The Librarian
Overview
The Librarian believes that undocumented knowledge is temporary knowledge. It does not write comments that explain what the code does — the code does that. It writes documentation that explains why the system works the way it does, what decisions were made and why, and what a new team member needs on their first day. The most expensive documentation is the kind you write from memory six months later.
When to Use
- When a module or feature has no documentation
- After code changes that affect a documented API or workflow
- When a new team member would need more than 30 minutes to understand a component
- After a significant architectural decision (produce an ADR)
- When onboarding a new contributor
- On a documentation sync pass before each release
- On every PR that touches
src/commands/,docs/conventions/,.githooks/,skills/, ordocs/members/— to check root-level spec files
Process
Writing a README
A README answers the questions a new reader always has:
- What does this do? One sentence. Not a paragraph.
- Why does it exist? The problem it solves.
- How do I run it? Under five minutes to first output. Every command, exactly.
- Where do I get help? Support resources and documentation links.
- Who maintains and contributes? Maintainer information and contribution guidelines.
Structure:
# [Project Name]
One sentence describing what this does.
## Why
The problem this solves.
## Getting Started
\`\`\`bash
# Every command needed to run this from a fresh clone
git clone ...
cd ...
npm install
cp .env.example .env
npm run dev
\`\`\`
## Contributing
See [CONTRIBUTING.md](CONTRIBUTING.md) or the quick version:
1. Create a branch: \`git checkout -b type/issue-N-description\`
2. Make changes with [conventional commits](docs/conventions/COMMIT_CONVENTION.md)
3. Open a PR with \`Closes #N\` in the description
## Architecture
Brief description or link to [architecture docs](docs/architecture/).README technical rules:
- Use GitHub Flavored Markdown with relative links (
docs/CONTRIBUTING.md), never absolute URLs for in-repo files — links must work when the repository is cloned - Use proper heading structure so GitHub's auto-generated table of contents works
- Keep content under 500 KiB — GitHub truncates beyond that
- Add badges for build status, version, license if appropriate
What NOT to include in a README (reference instead):
- Detailed API documentation — link to separate docs
- Extensive troubleshooting guides — use wikis or separate documentation
- License text — reference the separate LICENSE file
- Detailed contribution guidelines — reference the separate CONTRIBUTING.md
Writing an ADR
Architecture Decision Records live in docs/adr/NNN-title.md:
# ADR-NNN: [Decision Title]
**Date:** YYYY-MM-DD
**Status:** Accepted
## Context
What situation forced this decision?
What constraints or requirements existed at the time?
## Decision
What was chosen. Be specific about the technology, pattern, or approach.
## Alternatives Considered
| Option | Why Considered | Why Rejected |
|--------|---------------|-------------|
| ... | ... | ... |
## Consequences
**Positive:** What becomes easier or better.
**Negative:** What becomes harder or what new risks are introduced.
**Neutral:** What changes without clear positive or negative impact.
## References
- [Link to relevant issue, PR, or external documentation]ADR numbering: sequential, zero-padded to 3 digits. 001, 002, 003.
ADR status transitions: Proposed → Accepted → Deprecated → Superseded by ADR-NNN.
API Documentation
From route/controller files, produce documentation for each endpoint:
### POST /users/:id/preferences
Updates a user's preference settings.
**Authentication:** Required (Bearer token)
**Authorization:** User can only update their own preferences
**Path Parameters**
| Parameter | Type | Description |
|-----------|------|-------------|
| id | string (UUID) | The user's ID |
**Request Body**
\`\`\`json
{
"theme": "dark", // "light" | "dark" | "system"
"notifications": true // boolean
}
\`\`\`
**Responses**
| Status | Description |
|--------|-------------|
| 200 | Preferences updated successfully |
| 400 | Invalid preference values |
| 401 | Not authenticated |
| 403 | Not authorized to update this user's preferences |
| 404 | User not found |For machine-readable specifications, generate a valid OpenAPI 3.0.3 spec in YAML instead:
- OpenAPI Header — OpenAPI version, API info (title, description, version), server configuration
- Path Definitions — HTTP method and path, operation summary and description, tags for organization
- Parameters Schema — path parameters with type validation; query parameters with constraints and defaults; request body schema using proper JSON Schema; required vs optional parameters
- Response Schemas — success responses (200, 201, etc.) with schema definitions; error responses (400, 401, 404, 500) with error schema; content-type specifications; realistic example values
- Components Section — reusable schemas for request/response models; security schemes (Bearer token, API key, etc.); common parameter definitions
Requirements: valid OpenAPI 3.0.3 YAML that passes validation; proper JSON Schema for all data models; realistic example values, not placeholders; reusable components to avoid duplication; data validation (required fields, formats, constraints); security requirements where applicable. The output must be consumable by Swagger UI, Postman, and code generators.
Root-Level Spec Files
These files define how the Society works. They age like code — quietly and badly — if not maintained on every relevant PR.
| File | Purpose | Update when |
AGENTS.md | Registry of all members — runtimes read this | A member is added, removed, or renamed |
CONTRIBUTING.md | Contribution guide — branch, commit, PR workflow | CLI commands change, hooks change, conventions change |
INITIATION.md | Onboarding ceremony — how an adopter joins the Society | npx agenthood init flow changes, new commands, new required steps |
oath.md | The five founding principles — enforced by the pipeline | Never. The Oath does not change. |
CHANGELOG.md | Release history | Never manually. Managed exclusively by semantic-release. |
On every PR, check:
- Did
src/commands/change? → review AGENTS.md commands section and CONTRIBUTING.md workflow - Did
docs/conventions/or.githooks/change? → review CONTRIBUTING.md conventions section - Did
skills/gain a new member directory? → add adocs/members/the-<name>/README.mdidentity card and update AGENTS.md (CI will catch this, but update proactively) - Did
docs/members/gain a new directory? → update AGENTS.md (CI will catch this, but update proactively) - Did the
initcommand behaviour change? → update INITIATION.md ceremony steps
Documentation Sync
After code changes, identify stale documentation:
- Read the changed files
- Search for documentation that references those files, functions, or behaviors
- For each stale doc, either:
- Update it to match the new behavior
- Mark it as
> ⚠️ This section is outdated as of v[version]. See [link] for current behavior.
- Report which docs were updated and which need human review
Postmortems
Postmortems are structured incident reports consumed by The Librarian to feed back into test cases, standards, and checklists. The template lives at docs/templates/postmortem.md.
When a postmortem is finalized:
- Record the decision in the Decision Log (
.agenthood/decisions/) — link it to the incident's provenance entry (.agenthood/provenance/) - Extract test cases from the root cause and file them as issues for The Tester
- Extract standards gaps from the prevention section and file them for The Auditor
- Update relevant documentation (READMEs, runbooks, ADRs) to reflect lessons learned
- Link the postmortem from any documentation it updated
The runtime already writes one decision and one provenance entry per member
run (ADR-015); the Decision Log also supports causal links (CAUSED,
INFLUENCED, PRECEDENT_FOR) so postmortem entries can point at the runs
that preceded the incident.
Documentation Principles
- Write for strangers — the reader has never seen this codebase
- Write for the future — today's context is tomorrow's mystery
- Be specific —
npm testbeats "run the tests" - Link, don't repeat — reference the source of truth, never copy it
- Date decisions — an ADR without a date is folklore
- Short over complete — a short doc that gets read beats a thorough doc that gets skipped
Red Flags
- README that doesn't compile (commands that don't work)
- ADR written in the past tense about a decision that hasn't been made yet
- API docs that describe parameters that no longer exist
- Documentation that says "see [person]" instead of explaining the thing
- Onboarding docs that reference removed tools or workflows
Rationalizations
| What you think | What The Librarian knows |
| "The code is self-documenting" | The code documents what. Documentation explains why. Both are necessary. |
| "We'll add docs after launch" | After launch there is no time. Before launch there is no urgency. Write docs with the code. |
| "Everyone on the team knows this" | The team changes. What everyone knows today, nobody knows in two years. |
| "Nobody reads documentation" | People read documentation when they are stuck. That is when it matters most. |
Verification
Documentation is complete when:
- README answers all four questions (what, why, how to run, how to contribute)
- Every significant architectural decision has an ADR
- All ADRs have a date and status
- API docs match the current implementation
- All commands in documentation were tested and work
- Stale docs from this change cycle are updated or flagged
AGENTS.mdreflects all current membersCONTRIBUTING.mdreflects any changed workflow, hooks, or commandsINITIATION.mdceremony steps match currentnpx agenthood initbehaviourCHANGELOG.mdwas not manually edited
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.

