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/developerz-ai/ai-task-master/claude-mdgit clone --depth 1 https://github.com/developerz-ai/ai-task-masterWhat 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 | $0.02937 | $0.02937 |
| Opus 5 | $0.01469 | $0.01469 |
| Sonnet 5 | $0.00587 | $0.00587 |
| Haiku 4.5 | $0.00294 | $0.00294 |
Grade A, and why
ai-task-master CLAUDE.md scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- Use `node:fs/promises`, `node:child_process` / `execa`, web `fetch`. Avoid `Bun.file`, `Bun.$`, `Bun.spawn` in shipped code. They are fine in dev scripts and tests gated behind `if (process.versions.bun)`. How it starts
The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Instructions for Claude when editing aitm source. Not for end users.
House style
- SOLID. One responsibility per module. If a file grows a second reason to change, split it.
- Every module ships with tests. SRP + tested is the bar — no test, no merge.
- No premature abstraction. Inline first, extract on the second real caller.
- No legacy. When a design is replaced, the old one is DELETED — not kept behind a flag, an
optional field, a fallback branch, or a rename alias. A superseded path that still runs is worse
than no path: it is untested in anger, it drags its assumptions into new code, and it fires exactly
when something else already went wrong. If a failure needs a safety net, the net is "do less"
(skip the phase, let the caller proceed as it would have), never "run the thing we just rejected".
Two exceptions, both about DATA rather than code: an on-disk state/config file written by an older
version must still parse (
state/migrations), and a documented CLI/config key is renamed in one release with the old name removed, not aliased forever. - No comments unless the WHY is non-obvious. Names carry intent.
- Conventional commits. No co-author trailers.
Runtime stance
- Bun is the dev runtime —
bun run,bun test,bun.lockb. - Code must run unchanged on Bun, Node ≥ 20, and Deno ≥ 1.40. Treat Bun-only APIs as a portability bug.
- Use
node:fs/promises,node:child_process/execa, webfetch. AvoidBun.file,Bun.$,Bun.spawnin shipped code. They are fine in dev scripts and tests gated behindif (process.versions.bun). - Module system: ESM only.
"type": "module"inpackage.json.
TypeScript
strict: true. Noany, noas unknown as. Useunknownand narrow.typefor data shapes.interfaceonly when declaration merging is actually needed.- Named exports only. No default exports.
- File names: kebab-case. Type names: PascalCase. Function names: camelCase.
Provider
- Only OpenRouter (OpenAI-compatible) is used for inference. No Anthropic SDK, ever.
CredentialsreadsOPENROUTER_API_KEYfrom env. No OAuth, no~/.claude/.credentials.json.- The presence of
CLAUDE.mdin a target repo is a coding-style signal, not a provider signal — feed it to subagent system prompts.
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 · 142 lines · 2,937 tokens per session scan A 7d5415ce1b6f
ai-task-master CLAUDE.md is an instructions file published in the GitHub repository developerz-ai/ai-task-master (5 stars, last pushed 27d ago), licensed MIT. It adds 2,937 tokens to every session, about $0.0147 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 2026-08-31.
Other instructions, from other repositories
fff AGENTS.md
AGENTS.md instructions for dmtrKovalenko/fff, covering to clankers, development commands, building, testing and development tools and code quality.
claude_code_vs CLAUDE.md
Instructions for firish/claude_code_vs, covering claude.md, what this is, working agreement (how we collaborate here), architecture (where things live) and tech stack & hard constraints.
pi-coding-agent-forge AGENTS.md
Instructions for Firstp1ck/pi-coding-agent-forge, covering repository documentation rules, documentation goal, required documentation layers, readme.md — human guide and technical.md — advanced user reference.
GENesis-AGI AGENTS.md
AGENTS.md instructions for WingedGuardian/GENesis-AGI, covering agent instructions, code review mandate (adversarial), gitnexus — code intelligence (advisory), genesis capability surface and skills.
agentconfig.org AGENTS.md
Instructions for agentconfig/agentconfig.org, covering agent instructions for agentconfig.org, project overview, target audience, site structure and tech stack.
rosetta AGENTS.md
Instructions for tikoci/rosetta, covering codex instructions for rosetta, first reads, development defaults and mcp and client config.