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/notasithlord/peerd/agents-mdgit clone --depth 1 https://github.com/NotASithLord/peerdWhat 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 | $0.07969 | $0.07969 |
| Opus 5 | $0.03985 | $0.03985 |
| Sonnet 5 | $0.01594 | $0.01594 |
| Haiku 4.5 | $0.00797 | $0.00797 |
Grade A, and why
peerd 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 2d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- peerd CLAUDE.md — 91% identical, 72 lines differ
How it starts
The opening of the file, as written. The whole thing — 504 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — peerd
Persistent project context for Codex. Read this first on every session. Equivalent to OpenCode's
AGENTS.mdpattern.
What peerd is
A browser-native AI agent harness, shipped as a Chrome/Firefox extension. The agent runs entirely in the user's browser. It talks to a model API directly (BYOK — bring your own key), drives the browser's tabs and DOM, and can run shell commands in a sandboxed Linux VM compiled to WebAssembly. No backend, no telemetry, no account.
peerd is 0.x — experimental beta (breaking changes likely; no "V1"
commitment — versions stay 0.x until the surface stabilizes). The
initial feature buildout is complete and integrated; current work is
store-readiness polish and field hardening. Stack:
Manifest V3, Chrome and Firefox (via webextension-polyfill), vanilla
JS, no
build step.
The codebase is five modules
Each maps to one letter and color in the brand wordmark:
| Letter | Color | Module | Role |
|---|---|---|---|
p |
cyan | peerd-provider/ |
Model adapters (Anthropic, OpenRouter, OpenAI, Z.ai GLM, and keyless Ollama shipped; local WebGPU gated on the resident engine — registry.js is the live list) |
e |
red | peerd-egress/ |
Security: vault, allowlist (safeFetch), denylist, audit |
e |
amber | peerd-engine/ |
Execution instances — Sandboxes. Three kinds run in their own visible tab: WebVMs (CheerpX Linux), Notebooks (sealed JS worker + OPFS), Apps (opaque-origin iframe). A fourth, the headless worker (script), runs the Notebook's sealed worker offscreen with no tab — the agent's own quick compute. The sandbox is the isolate; a tab is one way to host it (taxonomy in the peerd-engine/ code). |
r |
green | peerd-runtime/ |
Agent loop, tools + per-environment actors (message_actor), sessions, profiles, skills, memory, permissions (Plan/Act), review, goal mode (autonomous loop), composer, cost, transfer, voice, clock, web tool policy |
d |
magenta | peerd-distributed/ |
The dweb. An always-on P2P base network (offscreen mesh + DHT + gossip), did:key identity, signed content addressing, the dwapp bridge, and a peer-to-peer app store that users AND the agent build, share, and run dwapps on. Chrome preview only until Firefox has a mesh host. |
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.
- 2d ago First seen · 504 lines · 7,969 tokens per session scan A d47658fda440
peerd AGENTS.md is an instructions file published in the GitHub repository NotASithLord/peerd (399 stars, last pushed 3d ago), licensed Apache-2.0. It adds 7,969 tokens to every session, about $0.0398 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
recap AGENTS.md
Instructions for atliq/recap, covering agents.md, what recap is, setup & run, pip - simple fallback and testing.
LearnLoop copilot-instructions.md
Instructions for Bogzx/LearnLoop, covering trailhead coaching — always on, never block, default order of operations on every code task, coach — call before answering any code task, skip detection and wikisave — call when the user states a teamwide convention.
byoky CLAUDE.md
Instructions for MichaelLod/byoky, covering byoky, project structure, development, architecture and conventions.
recap CLAUDE.md
Instructions for atliq/recap, covering claude.md and claude code.
vibe-annotations CLAUDE.md
Claude Code instructions for RaphaelRegnier/vibe-annotations, covering vibe annotations — monorepo, repo structure, quick commands, how the pieces relate and package-specific context.
multi-ai-panel AGENTS.md
AGENTS.md instructions for xiaojiou176-open/multi-ai-panel, covering agents, public surface rules, current visibility truth, protected paths and no-code-loss and destructive action discipline.