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.
npx agentmods add instructions/dralkh/seerai/agents-mdgit clone --depth 1 https://github.com/dralkh/seeraiWrote 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.
[](https://agentmods.dev/instructions/dralkh/seerai/agents-md)<a href="https://agentmods.dev/instructions/dralkh/seerai/agents-md"><img src="https://agentmods.dev/badge/instructions/dralkh/seerai/agents-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.11896 | $0.11896 |
| Opus 5 | $0.05948 | $0.05948 |
| Sonnet 5 | $0.02379 | $0.02379 |
| Haiku 4.5 | $0.01190 | $0.01190 |
Grade A, and why
seerai AGENTS.md scanned grade A with 1 finding 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 4d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- **`workspace_bash` behavior differs**: In the plugin, it prompts the user to run the command manually. In the MCP server, it executes real shell commands via `node:child_process` in the workspace directory. How it starts
The opening of the file, as written. The whole thing — 668 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Project Overview
seerai is an intelligent research assistant plugin for Zotero 8/9 that integrates AI-powered chat, federated scholarly search, RAG, OCR, systematic review, cloud storage, an agent skills library, and structured data extraction into the research workflow. The repo is a monorepo containing:
- Zotero Plugin (
src/,addon/) — runs inside Zotero's Firefox-based runtime (NOT Node.js) - MCP Server (
mcp-server/) — standalone Node.js server exposing Zotero tools via Model Context Protocol - Agent Skills Library (
skills/) — ~148 bundled, self-contained skill packages (SKILL.md + references/scripts) surfaced to agents via the skills tools
Commands
npm start # Dev server with hot reload (zotero-plugin serve)
npm run build # Production build: MCP bundle → plugin .xpi + tsc --noEmit + copies artifacts to root
npm run lint:check # Prettier check + ESLint
npm run lint:fix # Prettier write + ESLint fix
npm run release # Create GitHub release (zotero-plugin release)
npm run test # Run test suite (zotero-plugin test → mocha-based)
npm run test:extraction-eval # LLM extraction evaluation harness (tsx test/extractionModelEval.ts)
npm run test:pipeline-eval # Systematic-review pipeline evaluation harness (tsx test/pipelineEval.ts)
npm run update-deps # Update all dependencies to latest (npm update --save)
# MCP server (separate package)
cd mcp-server && npm run dev # Run MCP server with tsx
cd mcp-server && npm run bundle # Bundle MCP server to dist/seerai-mcp.cjs
cd mcp-server && npm run build # TypeScript compile (tsc)
cd mcp-server && npm run start # Run compiled MCP server
Always run npm run lint:check and npm run build after making changes.
Tests (npm test): Only run when the change touches testable logic (service helpers, state shapes, protocol/template flow, persistence schemas, bug fixes). Skip for pure formatting, comment, UI-style-only, or non-functional edits. When a test is warranted, run only the relevant file rather than the full suite when possible (e.g. npx mocha test/<file>.test.ts).
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.
- 4d ago First seen · 668 lines · 11,896 tokens per session scan A fabed8aeb3b7
seerai AGENTS.md is an instructions file published in the GitHub repository dralkh/seerai (77 stars, last pushed 1mo ago), licensed MIT. It adds 11,896 tokens to every session, about $0.0595 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
zotero-zotadata AGENTS.md
Instructions for ydeng11/zotero-zotadata, covering agents.md, role, technology constraints, typescript only and styling with tailwind css.
mktero AGENTS.md
AGENTS.md instructions for tenglvjun/mktero, covering agents.md, project overview, commands, repository map and architecture and runtime rules.
zotero-format-metadata AGENTS.md
AGENTS.md instructions for northword/zotero-format-metadata, covering agents.md, commands, architecture, lifecycle: bootstrap → sandbox globals → hooks and rule system (the core abstraction).
zotero-format-metadata CLAUDE.md
Claude Code instructions for northword/zotero-format-metadata, a project described as: Linter for Zotero. A plugin for Zotero to format item metadata. Shortcut to set title rich text; set journal abbreviations, university places, and item languages, etc; detect duplicate items.
ZoFiles CLAUDE.md
Instructions for X1AOX1A/ZoFiles, covering claude.md — development guide for zofiles, quick start, what is this, architecture overview and data flow.
zotero-watch-folder CLAUDE.md
Claude Code instructions for josesiqueira/zotero-watch-folder, covering claude.md, critical rules — must follow, responses, planning mode and edit / change mode.