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 manishiitg/llm-provider-mcp --skill delegate-coding-agentgit clone --depth 1 https://github.com/manishiitg/llm-provider-mcpWrote 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/manishiitg/llm-provider-mcp/delegate-coding-agent)<a href="https://agentmods.dev/skills/manishiitg/llm-provider-mcp/delegate-coding-agent"><img src="https://agentmods.dev/badge/skills/manishiitg/llm-provider-mcp/delegate-coding-agent.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 5.1 | $0.00052 | $0.00913 |
| Opus 5 | $0.00026 | $0.00456 |
| Sonnet 5 | $0.00010 | $0.00183 |
| Haiku 4.5 | $0.00005 | $0.00091 |
Grade A, and why
delegate-coding-agent 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 7d 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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Delegate Coding Agent
Choose the smallest capable model, delegate a precise task, and verify the result in the host session.
Choose a Route
Call list_coding_agents, then call list_coding_agent_models for the selected provider. Treat the live catalog and the user's preference as authoritative; model names and account availability can change.
| Route | Use for | Current examples |
|---|---|---|
| Powerful | Ambiguous architecture, cross-module debugging, security-sensitive work, difficult planning, or final review of high-risk changes | Codex gpt-5.5; Claude claude-opus-5; Pi google/gemini-3.1-pro-preview or zai/glm-5.3 |
| Balanced | Normal feature work, refactors, tool-heavy implementation, bug investigation, and tests | Codex gpt-5.4; Claude claude-sonnet-5; Cursor composer-2.5; Pi google/gemini-3.7-flash, minimax/MiniMax-M3, or moonshotai/kimi-k3 |
| Fast | Mechanical edits, focused UI iteration, simple test fixes, repository search, summaries, and other bounded work with objective checks | Codex gpt-5.3-codex-spark; Claude claude-haiku-4-5-20251001; Pi google/gemini-3.5-flash-lite |
Use these as routing hints, not benchmark rankings. Prefer auto when the user has no model preference and provider-native routing is more useful than a fixed model.
Delegate
- Decide whether delegation is worth the startup and verification overhead. Keep tiny tasks in the host session.
- Select a provider and model from the live tools. Use a powerful route when requirements or correctness are uncertain; use a balanced or fast route only when the task is bounded and independently verifiable.
- Give
delegate_coding_agentone clear outcome. Include the allowed scope, constraints, non-goals, relevant verification commands, and the expected response. - Pass the current trusted project root as
working_dirautomatically. Do not ask the user to enter it. - Record the returned
job_id. The delegation is asynchronous. - Call
get_coding_agent_jobonly afterpoll_after_seconds. Continue until the status iscompleted,failed,cancelled, ortimed_out. - Keep ordinary polling lightweight. When the user asks what is happening or progress appears stale, call
get_coding_agent_jobwithinclude_terminal_output: trueto receive the bounded plain-text tmux tail. Usetmux_capture_commandonly for deeper history. - Read the terminal result, inspect the actual diff, and run relevant tests before reporting completion.
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.
- 7d ago First seen · 56 lines · 52 tokens per session scan A f32849b1adf9
delegate-coding-agent is a skill published in the GitHub repository manishiitg/llm-provider-mcp (1 stars, last pushed today), licensed MIT. It adds 52 tokens to every session and 913 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 2026-08-31.
Other skills, from other repositories
coding-agents-farm
To orchestrate parallel coding-agent farms (Claude, Codex, Copilot, Gemini, etc.) on isolated git worktrees.
specgate-work-delivery
Use when implementing, resuming, verifying, or reworking an approved SpecGate work item or Context Pack, or when SpecGate status names the implementing agent.
specgate-work-preparation
Use when preparing a request or existing source documents for SpecGate approval, creating quick SpecGate work, publishing an artifact version, or repairing artifact readiness before implementation.
apex-workflow
Use for configurable, high-rigor app execution across repos. Selects the lightest safe workflow mode from an app profile, creates or uses a slice manifest, and routes product, tracker, code-intelligence, UI/UX, and verification gates without hard-coding one app's rules.
orcast
Use when orchestrating multiple agents in Orca: fan out parallel workers, pipeline work across stages, adversarially verify findings, run judge panels or loop-until-dry sweeps, or delegate a task and get the result back. You write a Plan definition in real JavaScript — a worker-workflow graph of worker, action, and…
bumpup-versions
Consistently bump and synchronize versions across all packages in the monorepo using pnpm. Use this to ensure the root and all workspace packages share the same version number.