pc-style/librarian/librarian
Claude Code and agent skill by pc-style/librarian
- community
Checking that the SKILL.md file is still in the repository.
Access the public Librarian research service for source-backed answers about GitHub repositories and public web pages. Use this skill whenever the user asks you to research a GitHub repo, inspect public source files, search commit history, compare repository diffs, read public docs/pages, or answer questions that need concise citations from repo/web evidence. This is especially useful when you do not have direct network access, when you want a dedicated research subagent, or when the user mentions Librarian.
Install
npx vskill@latest install pc-style/librarian/librarianSKILL.md
Librarian
Use the hosted Librarian service to get fast, source-backed findings about GitHub repositories and public web pages.
Service domain: https://librarian.pcstyle.dev
Librarian is modeled after Amp's librarian subagent: a constrained research worker with no local shell, filesystem, or secret access. In Amp, callers invoke librarian({ query: string }); this eve service accepts the same research task as the session message field.
The reverse-engineered Amp client exposes the Librarian tool names and client-side activity metadata, but not the server-side Librarian prompt text. Treat this service as a behavior-compatible research worker, not as a verbatim copy of Amp's private prompt.
Research tools
The service exposes the same GitHub and web tool surface Amp's Librarian uses internally:
- GitHub:
read_github,search_github,commit_search,list_directory_github,list_repositories,glob_github,diff - Web:
web_search,read_web_page
Librarian runs those tools silently and returns a concise, source-backed answer with citations.
Authentication
Production requests require a valid, unexpired Vercel OIDC token authorized for the Librarian deployment's project and environment. Arbitrary bearer strings are rejected. Send the token on every request:
Use LIBRARIAN_AUTH_TOKEN:
LIBRARIAN_API_KEY remains a deprecated alias in the script, but it must contain the same valid OIDC token. The script fails closed when neither variable is set.
Quick usage
Prefer the bundled script when shell access is available. Librarian performs live repo/web research, so the command may run for 5+ minutes on broad or tool-heavy questions. Use a generous command timeout and do not assume it is stuck just because it is quiet while the remote agent works.
Or from inside this skill directory:
The script reads LIBRARIAN_AUTH_TOKEN (or the deprecated LIBRARIAN_API_KEY alias) and fails when no token is configured.
When to use Librarian
Use Librarian for:
- Questions about public GitHub repositories: files, directories, symbols, docs, package setup, examples, tests, commits, and diffs.
- Public web research where the answer should cite pages or docs.
- Repo-plus-web questions, such as whether a public SDK can plausibly integrate into a public project.
- Requests where a user wants a direct answer with evidence, not a browsing transcript.
Do not use Librarian for:
- Local filesystem inspection. Librarian cannot see your local checkout.
- Secret or environment-variable inspection.
- Private GitHub data unless the deployed service has been configured with an authorized connector for that user/session.
- Tasks that require making changes to a repository; Librarian researches and answers.
How to ask good questions
Include exact identifiers when possible:
- GitHub repo:
owner/repoor a GitHub repository URL. - File paths, symbols, package names, error strings, branch/ref names, or docs URLs.
- The answer shape you want: “summarize with source paths,” “return concise evidence,” “compare these two refs,” etc.
Good examples:
API contract
The script uses eve's HTTP session API. Map Amp's librarian({ query }) call to { "message": "<query>" }:
POST /eve/v1/sessionwith JSON{ "message": "..." }.- Read
sessionIdfrom the response. GET /eve/v1/session/:sessionId/streamand collectmessage.completedevents.
Set these headers:
If implementing your own client, stream events are newline-delimited JSON. Return the last completed assistant message as Librarian's answer.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/librarian/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/pc-style/librarian/librarian)