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/velhud/patchbay/agents-mdgit clone --depth 1 https://github.com/velhud/PatchBayWhat 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 | $0.03873 | $0.03873 |
| Opus 5 | $0.01937 | $0.01937 |
| Sonnet 5 | $0.00775 | $0.00775 |
| Haiku 4.5 | $0.00387 | $0.00387 |
Grade A, and why
PatchBay 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 yesterday.
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 — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Project Purpose
This repository provides a hybrid ChatGPT-to-local-Codex bridge.
The product exposes a local Streamable HTTP MCP server so ChatGPT web/Pro or another MCP-compatible client can:
- inspect allowed local workspaces through bounded context tools;
- manage durable named Codex workers through natural-language briefs, reports, imported artifact context, isolated worktrees, and bounded peer-worker context;
- delegate larger work to local Codex CLI jobs;
- inspect async job status, structured results, session refs, worker reports, peer-worker context, imported artifact state, and worktree diffs;
- use
.ai-bridgehandoff artifacts; - coordinate multiple ChatGPT/MCP sessions on one shared local server with session-local tool modes, session-relative ownership flags, explicit takeover for cross-owner mutation, and per-repository mutation locks;
- optionally use direct edit, bash, or transcript-read power tools when explicitly enabled.
The repo still supports local maintainer workflows, but do not describe the app as only a maintainer utility. The public identity is now the broader ChatGPT-to-local-Codex platform.
Product Self-Knowledge
- Treat PatchBay as a ChatGPT-first local control plane: ChatGPT brings conversation, Projects, memory, generated artifacts, and coordination; local Codex brings the repository, git state, tools, credentials, and execution.
- Start docs/config/behavior changes by checking
README.md,QUICKSTART.md,docs/project/why-patchbay.md,docs/architecture/overview.md,docs/reference/public-tool-surface.md, anddocs/user/chatgpt-instructions.md. - Keep the app self-describing enough that a future coding agent can update configuration, docs, tool metadata, and examples from repository context without needing private oral history.
- Do not replace concrete setup steps with vague philosophy. Add the rationale, then keep the exact command, ChatGPT connector step, expected tool result, and verification command.
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.
- yesterday First seen · 165 lines · 3,873 tokens per session scan A d36478ba052e
PatchBay AGENTS.md is an instructions file published in the GitHub repository velhud/PatchBay (5 stars, last pushed 23d ago), licensed MIT. It adds 3,873 tokens to every session, about $0.0194 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 instructions, from other repositories
exploreyc CLAUDE.md
Instructions for KonstantinMB/exploreyc, covering claude.md, project overview, tech stack, development commands and backend dev server.
claude-code-adapter-fastapi CLAUDE.md
Instructions for wangfumin1/claude-code-adapter-fastapi, covering claude.md, project overview, development commands, basic commands and environment setup.
backchannel CLAUDE.md
Instructions for talberthoule/backchannel, covering claude.md, coordination record, current codebase snapshot, build & run and sentrux structural analysis.
backchannel AGENTS.md
Instructions for talberthoule/backchannel, covering agents.md, current codebase snapshot, build & run, sentrux structural analysis and docker compose (primary).
ReciboSplit CLAUDE.md
Claude Code instructions for Jejorm/ReciboSplit, covering claude.md, what recibosplit is, current status, stack and how to work on this project.
gptme AGENTS.md
Instructions for gptme/gptme, covering agent instructions for gptme, git workflow, code style, testing and project structure.