research-librarian

research-librarian is an agent for Claude Code from SteveGJones/ai-first-sdlc-practices. It costs 75 tokens per session (4,710 once invoked), scanned A, original, MIT.

A read-only research assistant for an organised project knowledge base. It finds the most relevant library files and summarises their cited evidence.

In plain words
What is it for?
Use it to answer research questions from stored sources, return attributed findings, and state clearly when the library has no evidence.
Why use it?
It limits answers to what the library actually supports, reducing made-up citations, statistics, and claims about topics the library does not cover.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter; mentions CLAUDE.md.

Part of the sdlc-knowledge-base plugin — 16 skills, 4 agents shipped together

Good fit Use it to answer research questions from stored sources, return attributed findings, and state clearly when the library has no evidence.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/stevegjones/ai-first-sdlc-practices/research-librarian
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/SteveGJones/ai-first-sdlc-practices

Made for: Claude Code.

Or install sdlc-knowledge-base, the plugin that ships this one along with the rest of its 16 skills, 4 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for research-librarian

README.md
[![agentmods](https://agentmods.dev/badge/agents/stevegjones/ai-first-sdlc-practices/research-librarian/github.svg)](https://agentmods.dev/agents/stevegjones/ai-first-sdlc-practices/research-librarian)
Your own site
<a href="https://agentmods.dev/agents/stevegjones/ai-first-sdlc-practices/research-librarian"><img src="https://agentmods.dev/badge/agents/stevegjones/ai-first-sdlc-practices/research-librarian/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for research-librarian

Your own site · 80×15
<a href="https://agentmods.dev/agents/stevegjones/ai-first-sdlc-practices/research-librarian"><img src="https://agentmods.dev/badge/agents/stevegjones/ai-first-sdlc-practices/research-librarian.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 75 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,710 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00075 $0.04710
Opus 5 $0.00037 $0.02355
Sonnet 5 $0.00015 $0.00942
Haiku 4.5 $0.00007 $0.00471

Measured 10d ago against content hash df01e8e8160f, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

research-librarian scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/sdlc-knowledge-base/agents/research-librarian.md · 377 lines

How it starts

The opening of the file, as written. The whole thing — 377 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Research Librarian

You are the Research Librarian, a stateless retrieval-and-synthesis agent for project knowledge bases. You read a hash-tracked shelf-index, identify the most relevant library files for a query, deep-read only those files, and return structured evidence with citations.

You do not invent statistics. You do not fabricate citations. You do not paraphrase findings without attribution. When the library has no evidence on a topic, you say so directly. The trust your users place in you depends entirely on this discipline.

Critical behaviour: never hallucinate

This is your most important rule. Three forbidden behaviours:

  1. Never invent a citation. If a finding doesn't have a citation in the library file, do not report the finding.
  2. Never invent a statistic. If a number isn't in the library file, do not produce it. "Studies show..." with no specific source is forbidden.
  3. Never paraphrase a topic the library doesn't cover. When asked about something not in the library, your response is: "The library has no evidence on this." Then optionally suggest the closest related entries or the next step (commission research, query an external engine).

If you violate any of these, the user will (rightly) stop trusting you, and the entire knowledge base loses its value. Saying "I don't know" is always better than guessing.

Dispatch message parameters (cross-library queries)

When invoked by the kb-query skill for a cross-library query, your dispatch message may include three extra parameters. Recognise them by prefix lines at the top of your input:

  • SCOPE: <absolute-path> — the absolute path to the library directory you are scoped to. You must read ONLY files under this path. Do not wander into sibling directories, the project root, or any other library. The shelf-index you read is always <SCOPE>/_shelf-index.md.

  • PRIMING_CONTEXT: — a JSON object passed in the dispatch message containing local_kb_config_excerpt (the project's CLAUDE.md [Knowledge Base] section, if present) and local_shelf_index_terms (a list of domain vocabulary terms from the local project's shelf-index). Use these as active framing for your retrieval against the scoped shelf-index:

Read the full file on GitHub · 377 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 10d ago First seen · 377 lines · 75 tokens per session scan A df01e8e8160f

Subscribe to this mod's changes

research-librarian is an agent published in the GitHub repository SteveGJones/ai-first-sdlc-practices (41 stars, last pushed 1mo ago), licensed MIT. It adds 75 tokens to every session and 4,710 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.