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/catiesgames/catclaw/claude-mdgit clone --depth 1 https://github.com/CatiesGames/catclawWrote 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/catiesgames/catclaw/claude-md)<a href="https://agentmods.dev/instructions/catiesgames/catclaw/claude-md"><img src="https://agentmods.dev/badge/instructions/catiesgames/catclaw/claude-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 | $0.21122 | $0.21122 |
| Opus 5 | $0.10561 | $0.10561 |
| Sonnet 5 | $0.04224 | $0.04224 |
| Haiku 4.5 | $0.02112 | $0.02112 |
Grade B, and why
catclaw CLAUDE.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 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
39. **Codex image generation is auto-on for every codex agent — via one `-c` flag + a SKILL note, NOT a per-agent config knob** — image generation (`gpt-image-2`) is a codex built-in: `codex_args::codex_args_from` uncond How it starts
The opening of the file, as written. The whole thing — 410 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CatClaw — Development Guide
Project Overview
Rust gateway for Claude Code CLI — multi-agent, multi-channel (Discord/Telegram/Slack/TUI), session management with tool approval system.
Architecture
Channel Adapters (Discord/Telegram/Slack) → MsgContext → MessageRouter → SessionManager → ClaudeHandle (claude -p subprocess)
↑
WS Server ← TUI/WebUI (GatewayClient)
MCP Server ← Claude CLI (tool calls)
Key Design Patterns
- Hot-reload: Config changes go through WS
config.set→ gateway applies immediately. Agent tool/approval changes go throughagents.reload_tools. - Shared state:
ConfigusesArc<RwLock<Config>>,AgentRegistryusesArc<RwLock<AgentRegistry>>,AdapterFilterusesVec<Arc<RwLock<AdapterFilter>>>. All hot-reloadable. - Approval system: PreToolUse hook → WS
approval.request→ broadcast to TUI + forward to origin channel → user approves/denies → hook receives result. - Session metadata: JSON in
metadatacolumn storesmodel,channel_id,sender_id. UseSessionRowhelper methods. - Channel adapters:
ChannelAdaptertrait withsend_approval()default method. Discord uses embed+buttons, Telegram uses inline keyboard, Slack uses Block Kit buttons. - Streaming:
ChannelCapabilities.streamingflag. Slack supports native AI streaming (chat.startStream/appendStream/stopStream). Adapters implement optionalsend_stream_start()/send_stream_append()/send_stream_stop()methods.
Critical Rules
RwLockReadGuard Across Await
NEVER hold a std::sync::RwLockReadGuard across an .await point — it's not Send and will fail to compile inside tokio::spawn. Always extract to a local variable first:
// WRONG: guard lives across await
if let Some(agent) = registry.read().unwrap().get(&id).cloned() {
do_something(agent).await; // guard still alive here!
}
// RIGHT: drop guard before await
let agent = registry.read().unwrap().get(&id).cloned();
if let Some(agent) = agent {
do_something(agent).await;
}
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 · 410 lines · 21,122 tokens per session scan B 9f041bf62fe7
catclaw CLAUDE.md is an instructions file published in the GitHub repository CatiesGames/catclaw (5 stars, last pushed 11d ago), licensed MIT. It adds 21,122 tokens to every session, about $0.1056 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
discoclaw CLAUDE.md
Instructions for DiscoClaw/discoclaw, covering discoclaw, safety, context loading (strict), identity and workspace file ownership.
discoclaw AGENTS.md
Instructions for DiscoClaw/discoclaw: This file exists so that non-Claude agents (e.g. OpenAI Codex) discover project conventions automatically. The source of truth is CLAUDE.md — read that file for project rules, dev workflow, context loading, and safety guidelines.
Oh-my-ClaudeClaw CLAUDE.md
Instructions for Kit4Some/Oh-my-ClaudeClaw, covering openclaw-cc v2 — autonomous ai harness, mandatory behavior — execute on every interaction, phase 0: context injection (before responding), phase 1: automatic task analysis and phase 2: execution with quality gates.
ai-agents security.instructions.md
Instructions for rjmurillo/ai-agents, covering security file rules, must, should, must not and references.
Cerebro AGENTS.md
Instructions for AgenticFirst/Cerebro, covering cerebro — agent context, read before touching integrations, project layout, conventions a fresh agent would otherwise miss and running locally.
ClawFleet CLAUDE.md
Instructions for clawfleet/ClawFleet, covering claude.md, project, architecture layers, clawsandbox (infrastructure) and clawfleet (product).