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 skills/firstp1ck/pi-coding-agent-forge/subagent-governancenpx skills add Firstp1ck/pi-coding-agent-forge --skill subagent-governancegit clone --depth 1 https://github.com/Firstp1ck/pi-coding-agent-forgeWrote 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/firstp1ck/pi-coding-agent-forge/subagent-governance)<a href="https://agentmods.dev/skills/firstp1ck/pi-coding-agent-forge/subagent-governance"><img src="https://agentmods.dev/badge/skills/firstp1ck/pi-coding-agent-forge/subagent-governance.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 | $0.00074 | $0.03567 |
| Opus 5 | $0.00037 | $0.01784 |
| Sonnet 5 | $0.00015 | $0.00713 |
| Haiku 4.5 | $0.00007 | $0.00357 |
Grade A, and why
subagent-governance 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 5d 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 — 189 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Subagent Governance
Use this skill in the parent orchestrator session before delegating work to child agents, before replacing a child that did not deliver, and when accepting delegated results. It decides which delegation shapes are allowed and which evidence is required. It is guidance, not a runtime guard: loading it does not install a package, enable an integration, change settings, or block a tool call.
Scope and Boundary
- This skill controls admissibility. Whether delegation is allowed at all, how many children of which kinds may be declared, who may write, what a worker must be told, what a handoff must return, when a replacement launch is legal, and how a reviewer finding is dispositioned.
- The harness's delegation documentation controls runtime mechanics. Tool schemas, action names, launch/monitor/wait syntax, context modes, agent and chain authoring, configuration, and error handling. In Pi, the installed
pi-subagentsskill remains canonical for those mechanics; this skill does not restate, replace, or override it, and it is not a second mechanics reference. - Parent-only. Do not inject or follow this skill inside a spawned child. A child does not own delegation, fanout, integration, or completion decisions. The one exception is a child that the parent explicitly assigned bounded fanout; it stays inside that assignment.
- Admissibility and runtime success are independent. An admissible shape can still fail at runtime, and no runtime capability, convenience command, or saved workflow makes a noncompliant shape admissible.
When to Use
Use when the parent is deciding whether to delegate, what to declare in a delegation request, how to keep concurrent work safe, how to react to a failed or partial delegated run, or what to do with reviewer output.
Do not use for:
- Launch syntax, tool arguments, action names, context modes, scheduling, or error messages — those belong to the harness's mechanics documentation.
- Creating, editing, disabling, or ejecting agents and saved workflows.
- Cost, usage, or status inspection with no pending delegation decision.
- Work the parent should simply do itself, such as a direct edit, a direct answer, or a single-file fix.
- Feature scope, plan structure, completion gates, or report obligations, which belong to the feature workflow policy in use.
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 189 lines · 74 tokens per session scan A 287d690eb980
subagent-governance is a skill published in the GitHub repository Firstp1ck/pi-coding-agent-forge (77 stars, last pushed yesterday), licensed MIT. It adds 74 tokens to every session and 3,567 once invoked, about $0.0004 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
Avant-Garde Frontend Architect
Expert frontend design and UI implementation with a focus on bespoke, visually striking, and minimalist interfaces. Trigger with 'ULTRATHINK' prefix for deep architectural analysis. Automatically detects and uses existing UI libraries (Shadcn UI, Radix, MUI) while applying unique styling. Includes a personalized…
FPEF Evidence Analyzer
Systematic Find-Prove-Evidence-Fix framework for complex system analysis and intervention. Use when debugging failures, investigating incidents, analyzing performance issues, or optimizing complex systems.
MCP Universal Manager
Auto-discovery and health monitoring for MCP servers across Claude Code, Factory Droid, Antigravity, Gemini, and other AI platforms. Use when managing MCP servers, monitoring tool integration, or debugging MCP connectivity issues.
council
Run a 5-adviser council to stress-test any decision. Five distinct lenses (Contrarian, First-Principles, Expansionist, Outsider, Executor) answer independently, then anonymously peer-review each other, and a Chairman synthesizes a single clear recommendation. Kills sycophancy and groupthink on high-stakes calls.
UltraPlan Autonomous Planner
Advanced autonomous planning system with viral automation, predictive support, and $100M ARR methodology. Use when creating strategic plans, scaling operations, or implementing autonomous business systems.
compound-engineering
Orchestration layer that coordinates Debt-Sentinel, Red Team Tribunal, and Spec-Lock into unified workflows. Use when you need comprehensive quality assurance with multiple validation stages. Provides safe-edit workflows, CI pipelines, enhancement pipelines, and engineering metrics dashboard.