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/watt-mind/coach/agents-mdgit clone --depth 1 https://github.com/watt-mind/coachWhat 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 | $0.02768 | $0.02768 |
| Opus 5 | $0.01384 | $0.01384 |
| Sonnet 5 | $0.00554 | $0.00554 |
| Haiku 4.5 | $0.00277 | $0.00277 |
Grade B, and why
coach AGENTS.md scanned grade B 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
- Postgres: `sudo pg_ctlcluster 16 main start` (cluster `16 main`, listens on `localhost:5432`). How it starts
The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Guidelines (AGENTS.md)
Primary context for AI agents (Claude Code, Gemini CLI, Cursor) working in this repository.
Project
Coach Watts — AI endurance coaching platform. Nuxt 3 + Nuxt UI, Vercel AI SDK, Prisma, Trigger.dev.
Issue tracking
All work is tracked in Linear, team key CW. Issue IDs look like CW-105.
Read docs/04-guides/issue-management.md before picking up any ticket. It defines the label taxonomy, workflow states, the AI-ready ticket template, and the concurrent-agent claim protocol.
Task state lives in Linear, never in git-tracked markdown. Files under docs/issues/ are the archive of already-resolved issues plus specs — do not treat them as a live queue and do not update their status as a way of tracking work.
Non-negotiables for agents
- One git worktree per ticket, created by
bin/worktree-up.sh. Never work on two tickets in the same checkout, and never hand-roll a worktree — no baregit worktree add, no hand-edited.env, no improvised port, nopnpm devon whatever port is free. See Worktrees.bin/worktree-up.sh CW-105 feat slug # worktree + branch + database + port + .env - Claim before you code. Set the ticket to
In Progress, assign it to yourself, addai:in-progress— then re-read it. If the assignee is not you, another agent won the race; take the next ticket. - Only touch the ticket's
Owned Paths. If the work requires files outside that set, stop and move the ticket toBlocked. - Never mark a ticket
Donewithout clean verification output. Run the ticket's Verification Command and post the output as a comment. - Blocked is a state, not a vibe. Missing credentials, ambiguity, or an external dependency → move to
Blockedand say what you need. Never leave a stalled ticket sitting inIn Progress. - File follow-ups in Linear. Bugs, risks, missing requirements, tech debt, or out-of-scope todos discovered while working → create a
CWissue inTriage(see issue-management §8). Do not expand the current ticket's scope to absorb them.
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 · 99 lines · 2,768 tokens per session scan B 22cd5402270a
coach AGENTS.md is an instructions file published in the GitHub repository watt-mind/coach (81 stars, last pushed 2d ago), licensed Apache-2.0. It adds 2,768 tokens to every session, about $0.0138 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
nightshift-os CLAUDE.md
Instructions for jaredlangleycc/nightshift-os, covering nightshiftos — working memory, who i'm working with, values + non-negotiables, current state and offer + icp.
product-coach AGENTS.md
AGENTS.md instructions for bobchao/product-coach, covering product coach — agent instructions, boot sequence (every session), directory layout, memory operations and check-in reminders（定期回顧提醒）.
product-coach CLAUDE.md
Claude Code instructions for bobchao/product-coach, a project described as: An AI Product Coach that asks, not tells.
life-designer-pro AGENTS.md
AGENTS.md instructions for wei-wei-hu/life-designer-pro: Use the repo-scoped skill at .agents/skills/life-designer-pro/SKILL.md for life-design, integrated career-and-life planning, ordinary-day analysis, multiple future paths, prototype planning, and blueprint or image-prompt generation.
life-designer-pro CLAUDE.md
Claude Code instructions for wei-wei-hu/life-designer-pro: This repository contains the life-designer-pro skill.
Intervals-MCP-Server-with-STRYD AGENTS.md
AGENTS.md instructions for methylone/Intervals-MCP-Server-with-STRYD, covering agents.md, what this project is, project conventions, build, run, test and when adding features.