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 instructions/edgehero/pi-dispatch/claude-mdgit clone --depth 1 https://github.com/edgehero/pi-dispatchWhat 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 | $0.02501 | $0.02501 |
| Opus 5 | $0.01251 | $0.01251 |
| Sonnet 5 | $0.00500 | $0.00500 |
| Haiku 4.5 | $0.00250 | $0.00250 |
Grade A, and why
pi-dispatch CLAUDE.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 3d 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working on pi-dispatch
Notes for an AI agent working in this repository. Short on purpose: specs/ is the authority, and this file
exists to stop you from having to discover the load-bearing parts by breaking them.
What this project is
pi-dispatch runs the pi coding agent as a service. It lives in the background and, on a cron schedule or a forge event (an issue, a comment, a pull request), opens a container that runs a flow against a repository, does the work, and shuts the container down. A panel shows the triggers, the history of every past run and the spend, and can turn the whole thing off.
pi has no queue, no concurrency control, no spend limit and, by its own README, no permission system. This project is exactly that missing operational layer and nothing else. When a change would make pi-dispatch smarter about what the agent does, it probably belongs in a skill or a flow, not here.
Two consequences worth internalising before you design anything:
- The container is the boundary. Isolation is built by the worker's own
docker runargv, so nothing an image contains can weaken it. - Money is the other boundary. Every gate that costs nothing runs before every gate that costs something, and a paid container starts only after all of them pass.
The shape
| Path | What it is |
|---|---|
worker/ |
the queue consumer, the CLI (pi-dispatch), the forge hosts, doctor, service installer |
receiver/ |
the always-on trigger edge (pi-dispatch-receiver): webhook routes and the poller |
admin/ |
the operator console, a pi extension (/dispatch), TypeScript, bundled to dist/ |
image/ |
the job image and the in-container runner that implements the exit-code protocol |
deploy/ |
service units, wrappers and the compose file; worker/deploy/ is the published mirror |
specs/ |
constitution, requirements, design, interfaces, open questions. The source of truth |
docs/ |
operator-facing reference, one file per feature or forge |
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.
- 3d ago First seen · 144 lines · 2,501 tokens per session scan A 556f4626115c
pi-dispatch CLAUDE.md is an instructions file published in the GitHub repository edgehero/pi-dispatch (163 stars, last pushed 3d ago), licensed MIT. It adds 2,501 tokens to every session, about $0.0125 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 instructions, from other repositories
Seo-Promt-Master GEMINI.md
Instructions for umutxyp/Seo-Promt-Master, covering seo prompt master, the short version, run the tool and ground rules.
platform-skills copilot-instructions.md
Instructions for nitinjain999/platform-skills, covering platform engineering instructions for github copilot, version: 1.40.0, source: https://github.com/nitinjain999/platform-skills, response contract and core principles.
ai-collaboration-notebook AGENTS.md
Instructions for roam-bit/ai-collaboration-notebook, covering agents.md · ai 协作错题本, 这是什么, 第一部分：协作协议（protocol.md）, 🚨 开工前 30 秒自检清单（每次新任务必跑） and 跟我协作的 17 条核心规则.
ExtremeRouter AGENTS.md
Instructions for rsalmn/ExtremeRouter, covering extremerouter — lead engineer agent, core philosophy, thinking process, critical thinking and coding standards.
Seo-Promt-Master AGENTS.md
Instructions for umutxyp/Seo-Promt-Master, covering agent instructions for this repository and before you open start.md.
kimi-responses-adapter AGENTS.md
AGENTS.md instructions for jianyun8023/kimi-responses-adapter, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.