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/gigleai/cavil-loop/agents-mdgit clone --depth 1 https://github.com/GigleAI/cavil-loopWrote 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/gigleai/cavil-loop/agents-md)<a href="https://agentmods.dev/instructions/gigleai/cavil-loop/agents-md"><img src="https://agentmods.dev/badge/instructions/gigleai/cavil-loop/agents-md.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 | $0.03041 | $0.03041 |
| Opus 5 | $0.01520 | $0.01520 |
| Sonnet 5 | $0.00608 | $0.00608 |
| Haiku 4.5 | $0.00304 | $0.00304 |
Grade A, and why
cavil-loop 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 3d 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 — 196 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
English · 中文
Quick context for agents (Claude Code et al.) and maintainers working in this repo. New people / agents read this first, then README.md for the external intro.
What this project is
GigleAI/cavil-loop is an Agent Skill — a feature package loaded by AI coding tools like Claude Code. It makes GitHub issue / PR comments the I/O of a local AI: a 60-second background poller on your machine finds whatever issue / PR is labeled pending/agent, spins up Claude Code locally, lets it work, push, reply, flip the label. Background in README.md.
Meta nature: this project develops itself (dogfooding). The issues / PRs of this repo run through its own workflow. Edit a script — the next dispatch of itself uses the new version.
Directory layout
.
├── README.md / README.zh.md ← External intro (what it is, how to use)
├── AGENTS.md / AGENTS.zh.md ← This file
├── CONTRIBUTING.md / .zh.md ← External contributor PR guide
├── SKILL.md / SKILL.zh.md ← Claude Code skill metadata (frontmatter + entry)
├── LICENSE ← MIT
├── setup.sh ← Bootstraps the daemon into a host project
├── coding-agent.config.example ← Config template (every field commented)
├── scripts/
│ ├── _lib.sh ← Common library: config load, log, has_claude_session, run_gh
│ ├── agent-poll.sh ← Main poller (called by systemd timer on Linux / launchd LaunchAgent on macOS)
│ ├── dispatch-new-issue.sh ← Dispatch a fresh issue
│ ├── dispatch-issue-comment.sh ← Dispatch on new issue comment
│ ├── dispatch-pr-comment.sh ← Dispatch on new PR comment
│ ├── seed-state.sh ← Initial seed of state.json
│ ├── create-worktree.sh ← Build worktree (injects worker identity)
│ ├── cleanup-issue.sh ← Post-merge cleanup (worktree / tmux / project hook)
│ └── session-log.sh ← View tmux pane history
├── prompts/
│ ├── new-issue.template.md ← Prompt for new-issue dispatch
│ ├── issue-comment.template.md ← Prompt for new issue comment
│ └── pr-comment.template.md ← Prompt for new PR comment
├── systemd/ ← Linux scheduler
│ ├── [email protected] ← User-scoped template service
│ └── [email protected]
├── launchd/ ← macOS scheduler
│ └── dev.luosky.coding-agent-work-loop.plist.template ← Generated per project by setup.sh
└── docs/
├── architecture.md / .zh.md ← Five-state label machine + session model + design FAQ
├── collaboration.md / .zh.md ← Multi-human + multi-agent workflows via label suffixes
├── persistence.md / .zh.md ← Where every artifact lives + retention + resume SOPs
├── security.md / .zh.md ← Security model + label discipline
└── operations.md / .zh.md ← Config / file layout / schedulers / troubleshooting / uninstall
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.
- 3d ago First seen · 196 lines · 3,041 tokens per session scan A cb6c3cbad358
cavil-loop AGENTS.md is an instructions file published in the GitHub repository GigleAI/cavil-loop (10 stars, last pushed 4d ago), licensed MIT. It adds 3,041 tokens to every session, about $0.0152 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 2026-08-31.
Other instructions, from other repositories
amp-acp AGENTS.md
Instructions for tao12345666333/amp-acp, covering agents.md, commands, architecture and code style.
skills CLAUDE.md
Instructions for mixedbread-ai/skills, covering mixedbread skills, repository structure, conventions, adding a new skill and testing a skill.
dev-workspace AGENTS.md
Instructions for iwe-org/dev-workspace, covering agent operating manual, start of every session, the operating loop, conventions and iwe basics.
tidyfactor-design AGENTS.md
Instructions for alwkala/tidyfactor-design, covering ⚡ skill & 24 modular slash commands (7 lifecycle stages) and critical architecture (non-negotiable).
growmos AGENTS.md
Instructions for codician-team/growmos: This repository keeps a knowledge graph in .growmos/ (entities, typed relations, provenance, profiles, a journal). It is the shared world model that survives context windows. Treat it as memory you read at the start of work and write to as you develop. Zero-config commands.
coach CLAUDE.md
Instructions for Andamio-Platform/coach, covering claude.md and claude code-specific notes.