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/yolo-labz/wa/agents-mdgit clone --depth 1 https://github.com/yolo-labz/waWrote 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/instructions/yolo-labz/wa/agents-md)<a href="https://agentmods.dev/instructions/yolo-labz/wa/agents-md"><img src="https://agentmods.dev/badge/instructions/yolo-labz/wa/agents-md.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 | $0.01519 | $0.01519 |
| Opus 5 | $0.00759 | $0.00759 |
| Sonnet 5 | $0.00304 | $0.00304 |
| Haiku 4.5 | $0.00152 | $0.00152 |
Grade A, and why
wa 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 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 — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — reviewer rules for yolo-labz/wa
Enforcement checklist for the Codex adversarial reviewer (and any agent
reviewing a PR in this repo). The why lives in CLAUDE.md
rules 26–32 and docs/reliability.md; this file is
what to flag. Each gate maps to a CLAUDE.md rule and is PR-blocking.
These rules are incident-driven: a wa-burocracy ~12 h silent inbound stall and
an undiagnosable REST dispatcher panic (2026-06-01), plus the PR #595
speculative-fix near-miss. Review as if a real person's WhatsApp is on the line —
because it is.
Daemon-reliability gates
Real-liveness — CLAUDE.md R26
- FLAG any health / readiness / liveness check that asserts only transport
state (
IsConnected(),connected:true, keepalive / ping round-trip) with no inbound-delivery / app-progress signal. Socket-up is a green lie under WhatsApp server-side device demotion (keepalive validates the socket, never auth; logout/demotion arrives as a stream error, not a ping failure). - REQUIRE connection-health surfaces to express soft-stale ("connected, no inbound past threshold") as a first-class state, derived from the last-inbound timestamp — not synthesised from the socket.
Watchdog / auto-heal — CLAUDE.md R27
- For code that auto-acts on a health signal, REQUIRE all of: edge-triggered
(
healthy→staletransition, not level), opt-in (explicit env/flag; detect-only default), cooldown-bounded, reversible (touches only the live socket — noLogout, nosession.dbmutation, no QR re-pair), and a synctest test proving "fires exactly once per edge + cooldown honoured." - REJECT a self-heal that mutates session/pairing state. REJECT level-triggered action loops (reconnect-storm risk).
- Auto-heal is permitted ONLY for a narrow / reversible / well-understood failure class. Anything else MUST be detect-and-alert — emit, let a human decide.
Diagnosable before fix — CLAUDE.md R28
- REQUIRE every
recover()to logstring(debug.Stack())plus method/context — not just the panic value. A recover that logs onlyfmt.Sprintf("%v", rec)is itself a defect (an undiagnosable crash in the recovery path). - ACCEPT a diagnosability-only PR (stack capture, no root fix) as a complete, shippable deliverable. Do not demand the root fix in the same PR.
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 · 98 lines · 1,519 tokens per session scan A 05c934c97656
wa AGENTS.md is an instructions file published in the GitHub repository yolo-labz/wa (0 stars, last pushed 6d ago), licensed Apache-2.0. It adds 1,519 tokens to every session, about $0.0076 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
pueue AGENTS.md
AGENTS.md instructions for Nukesor/pueue: This project does not approve of large-scale work done by LLMs.
RepoYeti AGENTS.md
AGENTS.md instructions for LunarWerxs/RepoYeti, covering agents.md, the one thesis, layout, before you push and daemon.
aidaemon CLAUDE.md
Instructions for davo20019/aidaemon, covering agents.md / claude.md, build & run, pre-commit checklist, git safety and release notes.
loope CLAUDE.md
Instructions for ngthluu/loope, covering claude.md and lessons learned.
trustcard AGENTS.md
Instructions for davidnichols-ops/trustcard, covering trustcard — repo knowledge for agents, what this is, commands, architecture map (lib/) and entry points.
binky AGENTS.md
AGENTS.md instructions for andreisilva1/binky, covering binky for agents, when to reach for it, install it as a tool, the tools and a typical loop.