delegate-skills is a toolkit for assigning coding tasks to separate agent command-line tools while keeping control of review and commits. It is for developers who want to organize installed implementers into lanes for work such as features, tests, or UI, then delegate tasks by lane or directly. The catalogue entries define the setup and delegation workflows.
Borrowing it
Nothing to install: this file belongs to amElnagdy/delegate-skills. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/amElnagdy/delegate-skills/master/AGENTS.mdgit clone --depth 1 https://github.com/amElnagdy/delegate-skillsWrote 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/amelnagdy/delegate-skills/agents-md)<a href="https://agentmods.dev/instructions/amelnagdy/delegate-skills/agents-md"><img src="https://agentmods.dev/badge/instructions/amelnagdy/delegate-skills/agents-md/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/instructions/amelnagdy/delegate-skills/agents-md"><img src="https://agentmods.dev/badge/instructions/amelnagdy/delegate-skills/agents-md.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 | $0.03976 | $0.03976 |
| Opus 5.5 | $0.01590 | $0.01590 |
| Sonnet 5.5 | $0.00795 | $0.00795 |
| Haiku 4.5 | $0.00398 | $0.00398 |
Grade A, and why
delegate-skills 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 13d 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 — 130 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working on delegate-skills
This repo is a Skills CLI package of delegation skills —
skills that let an orchestrating agent drive a separate CLI coding agent as an implementer, then review
and land the result. Eighteen implementer skills ship today: claude-delegate (Claude Code),
cline-delegate (Cline CLI), codex-delegate (OpenAI Codex), opencode-delegate (OpenCode),
agy-delegate (Google Antigravity), grok-delegate (Grok Build), kimi-delegate (Kimi Code),
muse-delegate (Muse Code),
qoder-delegate (Qoder CLI), vibe-delegate (Mistral Vibe), cursor-delegate (Cursor Agent CLI),
pi-delegate (Pi CLI), omp-delegate (Oh My Pi), aider-delegate (Aider), copilot-delegate (GitHub Copilot CLI),
warp-delegate (Warp Agent CLI), zcode-delegate (Z.AI ZCode), and commandcode-delegate (Command Code); siblings like
gemini-delegate can be added
later without renaming the repo. Two utility skills ship alongside them: delegate-setup (configure fleet lanes — setup only,
never dispatches) and codex-background (Codex waiting boundary around registered existing relays).
Vocabulary
One controlled vocabulary keeps the docs from drifting and stops edits (human or AI) from coining new jargon. Use these terms; don't invent synonyms.
| Use | For | Not |
|---|---|---|
| delegate / delegation | the activity, and this skill family | "relay" (as the activity), "hand-off", "offload" |
| orchestrator | the driving agent (Claude Code, …) | "controller", "driver" |
| implementer | the separate agent (Claude, Cline, Codex, OpenCode, Antigravity, Grok, Kimi, Muse, Qoder, Vibe, Cursor, Pi, Oh My Pi, Aider, Copilot, Warp, ZCode, Command Code) | "worker", "sub-agent", "executor" |
| brief | the self-contained task spec sent to the implementer | "task file", "the prompt", "the spec" |
| gates | the project's test/lint/build commands | "checks", "CI" |
| dispatch | sending the brief to the implementer | "fire off", "kick off" |
| land | commit the verified work yourself | — |
relay / relay.mjs |
the dispatch script only | never a category of skills |
| lane | a named fleet binding: implementer + optional dials (model, effort / variant, …) |
"route", "profile" |
| fleet | the user's set of lanes (which CLI handles which kind of work) | — |
setup skill / delegate-setup |
utility that discovers CLIs and writes the lane map after approval | a *-delegate skill |
exec, sandbox, resume, session, subagents |
Codex's own terms — use verbatim | don't paraphrase them; never use subagents as a generic synonym for implementer |
-p / --print, --yolo (--dangerously-skip-permissions), --permission-mode (standard/plan/auto-accept), --tools-all, --resume, --continue, --effort, --max-turns |
Command Code's own terms — use verbatim when discussing cmd |
never say Command Code has a sandbox, or a write-capable mode between withheld-tools and --yolo: --permission-mode auto-accept and --tools-all do not enable edits headlessly. Never call the binary "the cmd shell" |
run, agent (build/plan), session |
OpenCode's own terms — use verbatim | "sandbox" (OpenCode has no sandbox enum; autonomy is the agent) |
exec, sandbox, --json, --prompt-file, --session-id, --disable-approval, --disable-write, --disable-shell, --trust-workspace, --user-input-auto-resolve, --reasoning-effort, --provider (echo/meta) |
Muse's own terms — use verbatim when discussing muse |
never call --yolo the default or a permission mode; muse resume is TUI-only and is never invoked; --provider is relay-only, not a fleet dial |
project, conversation, model, permissions, sandbox, TUI, tasks, subagents |
Antigravity's own terms — use verbatim when discussing agy |
don't use subagents as a generic synonym for implementer |
session, sandbox (workspace/read-only/off), permission-mode, effort, streaming-json |
Grok Build's own terms — use verbatim when discussing grok |
don't paraphrase them |
session, --continue, model alias, auto permission mode, plan mode, --yolo |
Kimi Code's own terms — use verbatim when discussing kimi |
don't paraphrase them |
session, --continue, --resume, plan mode, --force, --sandbox (enabled/disabled), --trust, models |
Cursor Agent's own terms — use verbatim when discussing cursor-agent |
don't paraphrase them |
session, --continue, --resume, permission mode (acceptEdits/plan/bypassPermissions), sandbox, subagents, agent teams, background sessions |
Claude Code's own terms — use verbatim when discussing Claude | never use subagents as a generic synonym for implementer |
session, --json, -v (verbose), --auto-approve, --cwd, --model, --provider, --id (unsupported by the JSON relay), --plan, --data-dir / CLINE_SANDBOX (sandbox), -t/--timeout (CLI's own flag) |
Cline's own terms — use verbatim when discussing cline. The relay's --timeout watchdog is a different flag with the same spelling |
don't invent a Cline permission-mode enum |
session, -c, --resume, permission mode (default/accept_edits/auto/bypass_permissions/dont_ask/plan), print mode, stream-json, model, context window |
Qoder CLI's own terms — use verbatim when discussing Qoder | don't paraphrase them |
agent run, conversation / --conversation, run_id, run_url, --cwd, --output-format ndjson, --profile, --skill, --no-snapshot, run-cloud |
Warp Agent CLI's own terms — use verbatim when discussing oz |
never call oz "the warp CLI" in a launch context — warp is the interactive TUI and cannot be relayed; don't invent a Warp sandbox or permission-mode enum (oz agent run has neither) |
--prompt, --output (streaming/json/text), --agent (plan/accept-edits/auto-approve), --max-turns, --max-price, --max-tokens, --trust, --resume, --continue, --enabled-tools, --disabled-tools |
Mistral Vibe's own terms — use verbatim when discussing vibe |
don't invent a Vibe sandbox enum; --trust is not a permission mode |
session, --continue, --session, print mode, --mode json, tools, context files, project trust |
Pi's own terms — use verbatim when discussing pi |
don't paraphrase them |
session, --continue, --resume / --session, --print, --mode json, omp models, --thinking, approvalMode (always-ask/write/yolo), --yolo / --auto-approve, --tools, --no-extensions / --no-skills / --no-rules, --max-time, --profile |
Oh My Pi's own terms — use verbatim when discussing omp |
never call omp "pi" in a launch context — pi is a different CLI (pi-delegate); don't invent an omp sandbox enum; --yolo is tools.approvalMode: yolo, not a permission-mode enum |
--message-file, --yes-always, --suggest-shell-commands, --auto-commits/--dirty-commits, --dry-run, --edit-format, --architect, --file/--read, chat history |
Aider's own terms — use verbatim when discussing aider |
Aider has no sandbox, no permission modes, and no session ids; don't imply any. --file/--read scope the chat context — never call them a boundary |
session, -p/--prompt, --output-format json (JSONL), tools, --allow-all-tools, mode plan, --resume/--continue, --model, --effort, copilot login / env tokens, sandbox (experimental, MXC) |
GitHub Copilot CLI's own terms — use verbatim when discussing copilot |
don't paraphrase them |
mode (build/edit/plan/yolo), session (sess_…), goal (--target), --attach, app-server, plugins, skills |
ZCode's own terms — use verbatim when discussing zcode |
don't call mode a sandbox or a permission mode; never present build/edit as usable headlessly |
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.
- 13d ago Changed · +14 lines · +465 tokens per session 87d464135fa7
- 1mo ago First seen · 116 lines · 3,511 tokens per session scan A a4820ea99b7d
delegate-skills AGENTS.md is an instructions file published in the GitHub repository amElnagdy/delegate-skills (2,513 stars, last pushed 13d ago), licensed MIT. It adds 3,976 tokens to every session, about $0.0159 per session on Opus 5.5. 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
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.