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/vericle/heyclaw/agents-mdgit clone --depth 1 https://github.com/vericle/heyclawWhat 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 | $0.00787 | $0.00787 |
| Opus 5 | $0.00394 | $0.00394 |
| Sonnet 5 | $0.00157 | $0.00157 |
| Haiku 4.5 | $0.00079 | $0.00079 |
Grade A, and why
heyclaw 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 — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.
HeyClaw
Python 3.12 voice assistant powered by ElevenLabs, Gemini, Mem0, and MCP tools. Server code and tests live in heyclaw/; local audio and wake-word code lives in satellite/app/, with tests in satellite/tests/. shared/heyclaw_shared/ is the heyclaw-shared package that both components install as an editable path dependency; it holds the settings, logging, performance instrumentation, and ElevenLabs provider config they have in common. heyclaw/workspace/ contains the instructions, profile, and skills loaded by the assistant at runtime.
Configuration
- Requires Python 3.12,
uv, GNU Make,ngrok, andlsof. - Prepare
heyclaw/.envfromheyclaw/.env.example. It contains logging settings only. - Prepare
heyclaw/config.jsonfromheyclaw/config.example.json. It defines server providers, gateway, agent defaults, workspace, memory, and MCP servers. - Prepare
satellite/.envandsatellite/config.jsonfrom their example files. They define local logging, audio, wake word, and the ElevenLabs connection. - Never commit credentials from either component.
- The full development stack exposes the HTTP API on port 8000, the Speech Engine WebSocket on port 3001, and the ngrok tunnel configured by
gateway.publicWsUrl.
Unix/WSL Commands
Use only Makefile targets from the repository root:
make setup: installs Python 3.12 and dependencies for both components; eachuv syncalso installsshared/in editable mode.make backend: stops local project processes, then starts the backend, Speech Engine, and ngrok tunnel.make satellite: starts the interactive voice satellite; it requires microphone and audio devices configured insatellite/config.json.make kill: stops processes on the ports used by the project.make check: formats the code, applies Ruff fixes, and runs type checking. It may modify files.
After make setup, uv run --project satellite heyclaw-satellite onboard
creates or refreshes both component config.json files without replacing existing
values. It also recommends explicit PortAudio device indices; it is a CLI workflow,
not a Makefile target. Pass --update-audio to replace only the existing audio
indices and echo mode after the connected hardware changes; interactive terminals
show selectable lists powered by Questionary.
If the new hardware is not the Windows default, pass --input-device-index and/or
--output-device-index together with --update-audio after inspecting
heyclaw-satellite devices.
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 · 47 lines · 787 tokens per session scan A bd43ac4910a0
heyclaw AGENTS.md is an instructions file published in the GitHub repository vericle/heyclaw (17 stars, last pushed 22d ago), licensed MIT. It adds 787 tokens to every session, about $0.0039 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
claude-youtube-editor CLAUDE.md
Instructions for hassancs91/claude-youtube-editor, covering claude.md — claude-youtube-editor, layout, the example and conventions (hard rules).
claude-faceless-shorts-creator CLAUDE.md
Instructions for hassancs91/claude-faceless-shorts-creator, covering claude.md — claude-faceless-shorts-creator, layout and conventions (hard rules).
mcp-arr AGENTS.md
Instructions for aplaceforallmystuff/mcp-arr, covering agents.md - mcp-arr, tech stack, architecture, development commands and build.
Unity-AI-Animation CLAUDE.md
Instructions for IvanMurzak/Unity-AI-Animation, covering claude.md, what this is, build / run, critical invariants and find detail in.
SunoMCP copilot-instructions.md
Copilot instructions for AceDataCloud/SunoMCP, covering copilot instructions for sunomcp, sync from docs, project structure and development.
livekit_agent_simulator AGENTS.md
Instructions for quangdang46/livekit_agent_simulator, covering agents.md — livekit-agent-simulator, boundary, product rule: generic core, not fit-to-one-repo, product rule: no stubborn patches (defaults first) and product rule: no dead features (keep the surface lean).