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/jmrplens/gitlab-mcp-server/agents-mdgit clone --depth 1 https://github.com/jmrplens/gitlab-mcp-serverWhat 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 | $0.02932 | $0.02932 |
| Opus 5 | $0.01466 | $0.01466 |
| Sonnet 5 | $0.00586 | $0.00586 |
| Haiku 4.5 | $0.00293 | $0.00293 |
Grade A, and why
gitlab-mcp-server AGENTS.md 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 2d 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.
How it starts
The opening of the file, as written. The whole thing — 216 lines — stays where its author put it; the contents beside it link to each section on GitHub.
gitlab-mcp-server — Agent Quick Reference
MCP server in Go that exposes GitLab REST + GraphQL as MCP tools, resources, and prompts. Communicates via stdio (default) or HTTP. Catalog-first tool registration: 176 sub-packages under
internal/tools/, with three runtime surfaces (dynamicdefault,meta,individual).
Read first
| Where | Why |
|---|---|
CLAUDE.md |
Full project context, env vars, ADR list, agents/skills catalog |
.github/copilot-instructions.md |
Auto-loaded by VS Code Copilot; has the language policy, env var table, E2E recipes |
.github/instructions/*.md |
Auto-applied coding standards (go, MCP, OWASP, comments, code review) |
docs/development/tool-surfaces-and-action-core.md |
Surface ownership and catalog projection rules |
docs/development/adr |
Architectural Decision Records (catalog-first is ADR-0004) |
OpenCode-specific wiring (this file's agents, skills, paths) lives in
opencode.json + .opencode/agent/. The canonical agents and skills also
exist under .github/agents/ and .claude/skills/ for Copilot/Claude; the
opencode mirror is a thin frontmatter shim so they show up in the OpenCode
CLI.
Hard invariants
- Catalog-first registration. For ordinary GitLab API actions, add or
update domain-local
ActionSpecs and handlers. Do not add package-localRegisterToolsfunctions or ad hocmcp.AddToolcalls. The catalog ininternal/tools/action_catalog.goprojects everything into meta, dynamic,gitlab://tools, audits, LLM files, and individual tool surfaces. (ADR-0004.) - English-only artifacts. Every file committed to this repo must be in English, including doc comments, ADRs, branches, commits, and MCP tool descriptions. Chat with the developer in any language you want.
- Conventional commits.
feat:,fix:,docs:,test:,refactor:,chore:. Use thegit-commitskill (.claude/skills/git-commit/) for auto-detected type/scope. - Coverage target: maximum possible, 100% when feasible. CI enforces an 80% floor; in practice the team pushes per-package coverage to 100% unless there is a hard blocker (generated code, third-party interfaces). When you cannot reach 100% on a function, add a documented reason at the call site and prefer table-driven tests that exercise both happy and error paths.
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.
- 2d ago First seen · 216 lines · 2,932 tokens per session scan A d4f7312b3839
gitlab-mcp-server AGENTS.md is an instructions file published in the GitHub repository jmrplens/gitlab-mcp-server (31 stars, last pushed 3d ago), licensed MIT. It adds 2,932 tokens to every session, about $0.0147 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.
Other instructions, from other repositories
graymatter AGENTS.md
Instructions for angelnicolasc/graymatter, covering agents.md, your tools, when to call which, first call and identity.
graymatter CLAUDE.md
Instructions for angelnicolasc/graymatter, covering claude.md, a user asked you to set graymatter up, you are working in a project that already uses graymatter and you are contributing to graymatter itself.
syllago CLAUDE.md
Instructions for OpenScribbler/syllago, covering syllago, build and test, key conventions, architectural decisions and testing requirements.
mcp-gitlab AGENTS.md
Instructions for vish288/mcp-gitlab, covering mcp-gitlab — agent context, architecture, patterns, mcp compliance rules and tool annotations (mandatory).
the-matrix CLAUDE.md
Instructions for 0merUfuk/the-matrix, covering the-matrix — autonomous agent ecosystem, tool inventory, system topology, data flow and runtime selection.
mcp-gitlab GEMINI.md
Instructions for vish288/mcp-gitlab, covering mcp-gitlab — gemini cli extension context, tool categories, common workflows and notes.