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 skills add mhlaghari/stuntman --skill relaygit clone --depth 1 https://github.com/mhlaghari/stuntmanWrote 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/skills/mhlaghari/stuntman/relay)<a href="https://agentmods.dev/skills/mhlaghari/stuntman/relay"><img src="https://agentmods.dev/badge/skills/mhlaghari/stuntman/relay/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mhlaghari/stuntman/relay"><img src="https://agentmods.dev/badge/skills/mhlaghari/stuntman/relay.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00066 | $0.01035 |
| Opus 5 | $0.00033 | $0.00517 |
| Sonnet 5 | $0.00013 | $0.00207 |
| Haiku 4.5 | $0.00007 | $0.00103 |
Grade A, and why
relay scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Stuntman relay
Read host and tool setup. The host plans and reviews; an independent worker can execute a prepared spec while the host is capped. The host's quota and the worker's quota may be shared: a Codex worker using this Codex account does not provide extra capacity around its limit.
Inspect the current host's window
In Claude Code, run "$STUNTMAN_ROOT/bin/window". This uses Claude's live
usage endpoint, with five_hour_pct, blocked, seconds_until_reset,
resets_at, and seven_day_pct. If it returns an error, report that the
quota is unavailable; do not treat that as zero use.
In Codex, run "$STUNTMAN_ROOT/bin/codex-window". This reads local session
files only, never credentials or a model. It returns available, live: false,
snapshot_age_s, stale, and windows containing used_percent,
window_minutes, resets_at (Unix seconds), and expired.
- A Codex snapshot is an observation, not a live quota check. Missing windows,
available: false,stale: true, or elapsed resets mean current headroom is unknown. Use current host quota information if available, or explain that/statusis needed. Never promise zero use after a reset. - Treat a known window at 90% or higher as a reason to preserve a handoff. Check every reported window; a weekly cap can outlast the shorter reset.
- For a known cap, the relevant reset is the latest reset among the binding windows, not automatically the five-hour reset. If a binding window has no reset time, do not invent one.
Continue or checkpoint
While the host has headroom, complete the next planning or review unit using delegate: prepare the exact spec, inspect the worker's diff, and run verification yourself before advancing the queue.
As a cap approaches, write .stuntman/relay-state.json with:
- Host, task, acceptance criteria, and the current queue position.
- Worker backend, optional model pin, session ID (if issued), and process handle/PID while running. Keep credentials out of this file.
- Absolute project, spec, stdout, and stderr paths; baseline and verification commands; whether work is running, awaiting review, or awaiting execution.
- Observed quota timestamp, binding windows, and the exact next step.
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 Changed · +2 lines · -56 tokens per session 2051b6212328
- 9d ago First seen · 88 lines · 122 tokens per session scan A 497af054db80
relay is a skill published in the GitHub repository mhlaghari/stuntman (13 stars, last pushed yesterday), licensed MIT. It adds 66 tokens to every session and 1,035 once invoked, about $0.0003 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 skills, from other repositories
clawrouter
Hosted-gateway LLM router — save 84% on inference costs. A local proxy that forwards each request to the blockrun.ai gateway, which routes to the cheapest capable model across 76 models from OpenAI, Anthropic, Google, DeepSeek, xAI, Z.AI, and more. 7 free open-weight models included. Also exposes realtime market data…
surf
Use this skill — NOT browser or webfetch — for ALL Surf crypto-data calls. 83 endpoints at localhost:8402/v1/surf/ covering CEX/DEX markets, on-chain SQL over 80+ ClickHouse tables (Ethereum, Base, Arbitrum, BSC, TRON, HyperEVM, Tempo), 100M+ labeled wallets, prediction markets (Polymarket + Kalshi), social/CT…
phone
Verify phone numbers (carrier + SIM-swap fraud signals) and place AI-powered outbound voice calls via BlockRun's gateway (Twilio + Bland.ai). Trigger when the user asks to look up a number, check fraud risk, buy/rent a phone number, or place an AI voice call. Payment is automatic via x402 from the wallet.
imagegen
Generate or edit images via BlockRun's image API. Trigger when the user asks to generate, create, draw, make an image — or to edit, modify, change, or retouch an existing image.
polymarket-trading
Use when the user wants to actually PLACE, manage, or redeem bets on Polymarket (not just read odds — that's the blockrunpredexon data tools). Covers setup (deposit wallet, funding, approvals), buy/sell with confirm gating, positions, redeeming winnings, geoblock handling, and the end-to-end flow.
release
Use this skill for EVERY ClawRouter release. Enforces the full checklist — version sync, CHANGELOG, build, tests, npm publish, git tag, GitHub release. No step can be skipped.