A delegation workflow in which Claude plans and reviews while a separate headless coding worker performs the implementation.

Use it to send implementation tasks to supported workers such as Codex, opencode, Antigravity, Muse, or a local Claude proxy.

The planning agent does not need to write every change itself, and the resulting work can be checked and revised through review.

delegate — Skill for Claude Code by mhlaghari/stuntman

Install

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.

Any agent

npx skills add mhlaghari/stuntman --skill delegate

Clone the repo

git clone --depth 1 https://github.com/mhlaghari/stuntman

Made for: Claude Code.

Or install stuntman, the plugin that ships this one along with the rest of its 9 skills, 9 hooks.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/mhlaghari/stuntman/delegate/github.svg)](https://agentmods.dev/skills/mhlaghari/stuntman/delegate)

Your own site

<a href="https://agentmods.dev/skills/mhlaghari/stuntman/delegate"><img src="https://agentmods.dev/badge/skills/mhlaghari/stuntman/delegate/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/mhlaghari/stuntman/delegate"><img src="https://agentmods.dev/badge/skills/mhlaghari/stuntman/delegate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 76 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 2,684 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits

NVIDIA SkillSpector warn 7 Sept 2026

SkillSpector: 1 finding, up to medium

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

medium Rogue Agent · line 47
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.

How audits are shown

Origin original No closer match found in the catalogue.

Token cost

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	$0.00076	$0.02684
Opus 5.5	$0.00030	$0.01074
Sonnet 5.5	$0.00015	$0.00537
Haiku 4.5	$0.00008	$0.00268

Measured 1mo ago against content hash 7ed0d94c50de, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

delegate scanned grade A 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 1mo 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

`curl -s -m 2 http://localhost:8082/v1/models -H "x-api-key: freecc" -o /dev/null -w "%{http_code}"`.

skills/delegate/SKILL.md · 213 lines

How it starts

The opening of the file, as written. The whole thing — 213 lines — stays where its author put it; the contents beside it link to each section on GitHub.

stuntman: plan → execute → review

Read host and tool setup first. You, in Claude Code or Codex, are the architect and reviewer. A worker model (stunt) is the executor. Never implement the task yourself unless the worker fails twice.

Resolve the worker command

STUNT="$STUNTMAN_ROOT/bin/stunt"

Use "$STUNT" everywhere below. The worker backend is $STUNTMAN_WORKER (claude via local proxy — the default —, opencode, codex, agy, or muse).

If the user specifies a worker, set STUNTMAN_WORKER for every exec and resume call. Otherwise preserve their environment/default. Save the backend and any model pin with the session ID so review resumes the same provider. Running in Codex does not automatically select Codex as the worker.

Several workers: `STUNTMAN_WORKER_ORDER`

The user can list more than one worker, most preferred first, each with an optional model:

STUNTMAN_WORKER_ORDER="agy=gemini-3.8-flash-high opencode=opencode/big-pickle codex"

When it is set and the user named no worker for this task, spread the work across the list instead of sending everything to one backend:

One unit: the first listed worker that passes the probe below.
Several units that touch different files: give each to a different listed worker and run them in parallel.
Scouts (step 0): prefer a listed worker that is free or flat-rate; keep the strongest model for the hardest unit.
A worker fails its probe, or two feedback rounds: move the unit to the next listed worker once before you take it over (step 4).

Set STUNTMAN_WORKER on each call, plus STUNTMAN_MODEL when the entry has =model; an entry without one uses that backend's own default. Record which worker ran each unit.

Probe before a large delegation. Subscription workers run out without warning. Send a one-line prompt with the same STUNTMAN_WORKER / STUNTMAN_MODEL ("$STUNT" exec "Reply with the single word ok."). If it returns is_error or a quota message, pick another model or backend and tell the user which one you used.

Read the full file on GitHub · 213 lines

Changes

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.

1mo ago Changed · +56 lines · +6 tokens per session 7ed0d94c50de
2mo ago Changed · +12 lines · -76 tokens per session 0e498b525cf5
2mo ago First seen · 145 lines · 146 tokens per session scan A bfe38ff783bf

Subscribe to this mod's changes

delegate is a skill published in the GitHub repository mhlaghari/stuntman (16 stars, last pushed 20d ago), licensed MIT. It adds 76 tokens to every session and 2,684 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Building agents Claude Code Cost Optimization ai-agents claude deepseek llm-orchestration

delegate next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
delegatemhlaghari/stuntman	Skill	16	20d ago	MIT	A	76	Claude Code
clawrouterBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	171	Claude Code, Codex
imagegenBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	45	Claude Code, Codex
phoneBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	72	Claude Code, Codex
polymarket-tradingBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	76	Claude Code, Codex
releaseBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	44	Claude Code, Codex
predexonBlockRunAI/ClawRouter	Skill	6.7k	today	MIT	A	84	Claude Code, Codex

delegate