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/harness/mcp-server/agents-mdgit clone --depth 1 https://github.com/harness/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 | $0.01942 | $0.01942 |
| Opus 5 | $0.00971 | $0.00971 |
| Sonnet 5 | $0.00388 | $0.00388 |
| Haiku 4.5 | $0.00194 | $0.00194 |
Grade A, and why
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 — 162 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Harness.io MCP Server
Registry-based MCP server: 11 consolidated tools (harness_list, harness_get, harness_create, harness_update, harness_delete, harness_execute, harness_search, harness_describe, harness_diagnose, harness_status, harness_schema) dispatching across 219+ resource types. TypeScript (Node.js 20+), ESM, pnpm, Zod v4.
Commands
pnpm install # install deps
pnpm build # tsc → build/
pnpm dev # tsc --watch
pnpm typecheck # tsc --noEmit
pnpm test # vitest run (118 test files, 2510 tests, ~9s)
pnpm standards:check # run coding-standards and structural validation tests
pnpm start # stdio transport (requires HARNESS_API_KEY)
pnpm start:http # HTTP transport (node build/index.js http)
pnpm inspect # MCP Inspector against stdio build
pnpm docs:generate # generate docs from build/ — run pnpm build first
pnpm docs:check # CI check (reads build/); stale build = wrong count and CI fail
Project Structure
src/
├── index.ts # Entrypoint — stdio/HTTP transport setup
├── config.ts # Env var validation (Zod); see .env.example for all vars
├── registry/
│ ├── index.ts # Core dispatch (~50K) — ResourceDefinition, EndpointSpec
│ ├── types.ts # Registry types: ResourceDefinition, EndpointSpec, OperationPolicy
│ ├── extractors.ts # Response shape extractors for all resource types
│ ├── scope-utils.ts # Scope resolution utilities
│ └── toolsets/ # 41 declarative toolset definition files (pure data — add new resources here)
├── tools/ # 11 consolidated tool handlers
├── prompts/ # ~33 prompt templates (one per workflow)
├── resources/ # MCP resources (pipeline YAML, execution summary)
├── client/
│ └── harness-client.ts # HTTP client — auth, retry, pagination
├── data/ # Example data and JSON schemas for entity validation
├── search/ # Cross-resource keyword search
├── audit/ # Audit manager
└── utils/ # Errors, logger, elicitation, deep-links, body normalizer
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 · 162 lines · 1,942 tokens per session scan A f85d0feb36c6
mcp-server AGENTS.md is an instructions file published in the GitHub repository harness/mcp-server (93 stars, last pushed 2d ago), licensed MIT. It adds 1,942 tokens to every session, about $0.0097 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
open-code-review AGENTS.md
Instructions for alibaba/open-code-review, covering agent guidelines for open-code-review, project overview, git commit notes, license headers and code style.
pydantic-ai AGENTS.md
Instructions for pydantic/pydantic-ai, covering your primary responsibility is to the project and its users, gathering context on the task, ensuring the task is ready for implementation, philosophy and requirements of all contributions.
ratel AGENTS.md
Instructions for ratel-ai/ratel, covering agents.md — working in the ratel repo, build & test, rust, ts and python (from src/sdk/python/; needs uv).
Trellis AGENTS.md
AGENTS.md instructions for mindfold-ai/Trellis, covering trellis instructions, gitnexus — code intelligence, always do, never do and resources.
sofagent GEMINI.md
Instructions for KongFangXun/sofagent, covering gemini.md · sofagent gemini cli 适配(薄挂载), 会话开始时按序加载(四层加载链), 审计强制(平台无关) and 连接 mcp server.
STAR AGENTS.md
AGENTS.md instructions for wanghao9610/STAR, covering agent instructions, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.