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 agents/leo-maomao/pm-copilot/agent-operating-modelgit clone --depth 1 https://github.com/Leo-maomao/pm-copilotWrote 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/agents/leo-maomao/pm-copilot/agent-operating-model)<a href="https://agentmods.dev/agents/leo-maomao/pm-copilot/agent-operating-model"><img src="https://agentmods.dev/badge/agents/leo-maomao/pm-copilot/agent-operating-model.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 | $0.00000 | $0.03526 |
| Opus 5 | $0.00000 | $0.01763 |
| Sonnet 5 | $0.00000 | $0.00705 |
| Haiku 4.5 | $0.00000 | $0.00353 |
Grade A, and why
agent-operating-model 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 4d 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 — 238 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Operating Model
PM Copilot is an AI Product Manager Agent System. The workflow is the safety rail; the agent experience is goal-driven product work.
Execution Boundary
This repository defines specialist roles, prompts, artifacts, bounded Loop decisions, and a local runtime adapter. It never embeds credentials or API keys. The adapter automatically reuses the user's active authenticated Seawork, Codex CLI, Claude CLI, Qwen Code, Kimi Code, Qoder CLI, or CodeBuddy Code session and model; it does not impose a fixed default runtime.
The PM Orchestrator creates specialist work only for independent, bounded questions and records the owned question, model/runtime, evidence delta, and reconciliation result. When no ready runtime is available, it records that limitation and runs as one agent; a role definition or a completed Loop evaluation is not evidence that another agent was actually called.
First-Principles Execution Rule
Every run starts from the outcome the user needs, not from an available skill, familiar workflow, artifact template, or tool. Decompose the work in this order:
Goal -> necessary conditions -> current blocking condition -> most direct next action
Keep decomposing until the next action has a direct causal contribution to removing the current blocker. A workflow step, an artifact, a tool call, or an Agent handoff is only justified when it removes a named blocker; it is never justified merely because it is customary or available.
Before Act, record or be ready to state:
- the user outcome and the condition that makes it true;
- the current blocker or unknown preventing that condition;
- why the chosen next action is the most direct safe way to remove it;
- what observation, answer, artifact, or validation result proves the blocker changed.
If this chain cannot be stated, do not continue by default. Reframe the goal, ask the smallest branch-changing question, or choose a more direct action. Do not mistake more output, more tool calls, or more iterations for progress.
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.
- 4d ago First seen · 238 lines · 0 tokens per session scan A 0d7488c1469e
agent-operating-model is an agent published in the GitHub repository Leo-maomao/pm-copilot (4 stars, last pushed today), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,526 tokens. 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 agents, from other repositories
archon
Autonomous vision agent. Decomposes vague or specific direction into campaign phases. Delegates to Marshals and specialists. Reviews output against quality standards. Maintains campaign state across invocations. Does not write code — orchestrates those who do.
implementer
PlanGate exec フェーズのタスク実装エージェント。workflow-conductor から委譲されたタスクを TDD で実装し、完了報告する。タスクごとに新規起動される使い捨てエージェント。.
orchestrator
PlanGate × Workflow/Skill/Agent ハイブリッドアーキテクチャの実行層総責任者。WF-01〜WF-05 の phase 遷移を制御し、各 phase の完了条件判定・Agent への委譲・handoff 発行を行う。汎用マルチエージェント調整も兼ねる。.
graph-planner
Analyze a coding request and design an executable dependency graph. Read-only.
ss-reviewer
Independent code reviewer for the specship review stage — one member of the opt-in review panel. Use to get fresh, context-free eyes on a working diff, optionally through a named lens (correctness, security, performance, contract-consistency). Reports findings only; it does not fix code, decide the verdict, or touch…
implementation-agent
確定した設計(design artifact)に従って 最小単位でコードを実装 する。以下を守る.