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/tonydzi/claw-consensus/agents-mdgit clone --depth 1 https://github.com/tonydzi/claw-consensusWrote 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/tonydzi/claw-consensus/agents-md)<a href="https://agentmods.dev/instructions/tonydzi/claw-consensus/agents-md"><img src="https://agentmods.dev/badge/instructions/tonydzi/claw-consensus/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 | $0.00979 | $0.00979 |
| Opus 5 | $0.00490 | $0.00490 |
| Sonnet 5 | $0.00196 | $0.00196 |
| Haiku 4.5 | $0.00098 | $0.00098 |
Grade A, and why
claw-consensus 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 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 — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — working in this repo
Written for AI coding agents, and equally readable by a human contributor. Short on purpose.
What this repo is
The reference implementation of a multi-machine coordination layer: a dual-rail message bus, ACK discipline, and a consensus protocol (propose → counter → accept → commit → verify) with an append-only ledger. It is a sanitized extract of a system that runs daily across several real machines; the personal content is gone, the protocol and the discipline are what's published.
Read docs/PROTOCOL.md before changing anything under reference/. The code
is small; the invariants are the hard part, and they are written down.
Stack and layout
- Python, stdlib only. No packages, by design: this has to run on a family laptop with a bad connection. Adding a dependency is a design change, not a convenience.
reference/consensus.py— the engine. Append-only JSONL ledger.reference/machine_bus.py— file-rail mailbox, capability addressing, auto-failover.reference/bus_send.py— the single entry point for machine-to-machine messages. Both rails fire from here, which is what makes a single-rail send impossible rather than discouraged.reference/sync_monitor.py,reference/fleet_sign.py,reference/protocol_guards.py.demo/demo.py— five self-checking offline scenarios.docs/FAILURE-MODES.md— every known failure, its root cause, and the guard that closed it.
How to verify a change
python demo/demo.py # all five scenarios, offline, self-checking
python demo/demo.py --json out.json # machine-readable timings
No network, no chat rail, no model, no packages — the demo sets CONSENSUS_NO_BUS=1 and simulates
two machines by flipping MACHINE_KEY. If your change is real, it shows up here. Paste the output
in the PR.
Scenario E exists to prove a corrupt line in a shard does not eat the events after it. If you touch ledger parsing, that scenario is the one that must stay green.
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 · 75 lines · 979 tokens per session scan A 90ef94fae427
claw-consensus AGENTS.md is an instructions file published in the GitHub repository tonydzi/claw-consensus (2 stars, last pushed 5d ago), licensed MIT. It adds 979 tokens to every session, about $0.0049 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
pact AGENTS.md
AGENTS.md instructions for TailorAU/pact, covering pact repo — agent onboarding, you are here, branch model, source-of-truth note (important) and open work (as of 2026-07-24).
gemot CLAUDE.md
Claude Code instructions for justinstimatze/gemot, covering gemot, key commands and architecture.
tinysdlc AGENTS.md
Instructions for Minh-Tam-Solution/tinysdlc, covering setup activity, team communication, single teammate, multiple teammates (parallel fan-out) and shared context.
writbase CLAUDE.md
Instructions for Writbase/writbase, covering writbase task context, task tracking (mandatory when doing task-related work) and agent team — writbase improvements.
LoopAgentTeams CLAUDE.md
Instructions for ouob-tw/LoopAgentTeams, a project described as: Spec-driven multi-agent teamwork for any skills-capable agent (Claude Code, Codex, and more) — plan, dispatch, monitor, QA. No framework lock-in.
fulcrum CLAUDE.md
Instructions for moabualruz/fulcrum, a project described as: Local-first Agent OS — project management, agent orchestration, docs, memory, and developer tooling. Runs on PGlite. SaaS-ready.