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/mcp-best-practicesgit clone --depth 1 https://github.com/jmrplens/gitlab-mcp-serverWrote 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/jmrplens/gitlab-mcp-server/mcp-best-practices)<a href="https://agentmods.dev/instructions/jmrplens/gitlab-mcp-server/mcp-best-practices"><img src="https://agentmods.dev/badge/instructions/jmrplens/gitlab-mcp-server/mcp-best-practices.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 5.1 | $0.01183 | $0.01183 |
| Opus 5 | $0.00592 | $0.00592 |
| Sonnet 5 | $0.00237 | $0.00237 |
| Haiku 4.5 | $0.00118 | $0.00118 |
Grade A, and why
gitlab-mcp-server mcp-best-practices.instructions.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 5d 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 — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MCP Best Practices
Protocol-level guidelines for building high-quality MCP servers that enable LLMs to accomplish real-world tasks effectively.
Tool Design Principles
Naming
- snake_case with service prefix:
gitlab_create_issue,gitlab_list_projects - Action-oriented verbs: get, list, search, create, update, delete
- Specific names: Avoid generic names that conflict with other MCP servers
- Descriptions must match behavior exactly: Narrow, unambiguous descriptions
Annotations
Every tool MUST declare annotations to help clients understand behavior:
Annotations: &mcp.ToolAnnotations{
ReadOnlyHint: boolPtr(true),
DestructiveHint: boolPtr(false),
IdempotentHint: boolPtr(true),
OpenWorldHint: boolPtr(true),
}
| Tool Type | ReadOnly | Destructive | Idempotent | OpenWorld |
|---|---|---|---|---|
| list / get / search | true | false | true | true |
| create | false | false | false | true |
| update | false | false | true | true |
| delete | false | true | true | true |
Atomic Operations
Keep tools focused on a single operation. Avoid multi-step tools that combine unrelated actions.
Response Formats
- JSON: Structured data for programmatic processing (include all fields)
- Markdown: Human-readable display (use headers, lists, formatting)
- Convert timestamps to human-readable format in Markdown
- Show display names with IDs in parentheses when useful
- Omit verbose metadata in Markdown output
Pagination
For list operations:
- Default to 20-50 items per page
- Always respect the
per_page/limitparameter - Return metadata:
has_more,total_items,page,total_pages - Never load all results into memory for large datasets
- Use cursor-based pagination when offset-based becomes expensive
Error Handling
- Actionable messages: Tell the LLM what went wrong AND what to try next
- Report tool errors in results: Use
IsError: trueinCallToolResultfor recoverable errors - Don't expose internals: Hide stack traces, internal paths, database details
- Wrap with context:
fmt.Errorf("gitlab_list_projects: %w", err) - Clean up resources: Always release connections, files on errors
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.
- 5d ago First seen · 143 lines · 1,183 tokens per session scan A 9114126497da
gitlab-mcp-server mcp-best-practices.instructions.md is an instructions file published in the GitHub repository jmrplens/gitlab-mcp-server (33 stars, last pushed today), licensed MIT. It adds 1,183 tokens to every session, about $0.0059 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
ollama AGENTS.md
AGENTS.md instructions for ollama/ollama, covering agents.md and building.
ollama CLAUDE.md
Claude Code instructions for ollama/ollama: See AGENTS.md for the shared agent instructions for this repository.
claude-tap AGENTS.md
Instructions for liaohch3/claude-tap, covering maintainer automation notes, agents 索引, documentation boundary, review guidelines and pre-commit hook.
opencodex AGENTS.md
AGENTS.md instructions for lidge-jun/opencodex, covering agents.md, what this project is, repository layout, optional subsystems stay off the core path and the devlog directory.
goai AGENTS.md
AGENTS.md instructions for zendev-sh/goai, covering agents.md - goai, commands, architecture, key rules and adding providers.
gollem CLAUDE.md
Instructions for gollem-dev/gollem, covering claude.md, restriction & rules, commands, development and testing and test execution.