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/marciommc/warden/agents-mdgit clone --depth 1 https://github.com/Marciommc/wardenWrote 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/marciommc/warden/agents-md)<a href="https://agentmods.dev/instructions/marciommc/warden/agents-md"><img src="https://agentmods.dev/badge/instructions/marciommc/warden/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 | $0.00762 | $0.00762 |
| Opus 5 | $0.00381 | $0.00381 |
| Sonnet 5 | $0.00152 | $0.00152 |
| Haiku 4.5 | $0.00076 | $0.00076 |
Grade A, and why
warden 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Warden
What this project is
Warden is a constitutional runtime for AI agents, exposed as an MCP server. Thesis: agents cannot mint their own authority. Identity is granted (atomic reservation), work moves in typed envelopes with loud rejection, promotion is human-only (the approval capability does not exist agent-side), and history is an append-only hash-chained ledger that cannot be rewritten silently.
Read README.md first — it is the constitution. Every change must strengthen
or demonstrate one of the four Articles. If a change weakens an Article, stop
and ask.
Hard rules
- Never add an MCP tool that approves, promotes, or closes work. The absence of that tool is the core argument of the project.
- Never make the ledger mutable, truncatable, or optional. Violations stay on the record forever.
- Rejections must be explicit and descriptive — echo what was attempted, what is allowed, and that the attempt was recorded. No silent failures, no coerced defaults.
- Zero heavy dependencies. Current stack: Node 22+, ESM, @modelcontextprotocol/sdk, zod. State = JSON file; ledger = JSONL. Do not introduce databases, frameworks, or build steps without an explicit request.
- Do not reference private companies, clients, internal ecosystems, or internal task systems in code, comments, docs, or commit messages. Public tools and interoperability targets may be named only when materially relevant to Warden. This is a standalone personal open-source project (Apache-2.0).
Layout
src/ledger.mjs— hash-chained append-only ledger (Article IV)src/warden.mjs— core primitives and violation classes (Articles I–III)src/server.mjs— MCP stdio server; agent-facing tools onlybin/warden-cli.mjs— human-side gate + replay (never reachable via MCP)demo/cheat.mjs— executable demo: every violation class, then the honest path, then ledger tampering caught by chain verification
Conventions
- Small, single-purpose commits with descriptive messages; commit after every working increment (commit history is part of the hackathon evidence).
- Any new violation class gets: a rule name in SCREAMING_SNAKE, a ledger event,
a loud rejection message, and coverage in
demo/cheat.mjs. - Tests: plain
nodescripts undertest/(no framework).npm testmust pass before any commit. - Keep tool descriptions in
server.mjsnormative — they tell the agent what is forbidden and that violations are recorded. They are part of the constitution, not decoration.
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 · 73 lines · 762 tokens per session scan A c976b57e7780
warden AGENTS.md is an instructions file published in the GitHub repository Marciommc/warden (0 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 762 tokens to every session, about $0.0038 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
agent-governance-toolkit copilot-instructions.md
Copilot instructions for microsoft/agent-governance-toolkit, covering copilot instructions for agent-governance-toolkit, architecture, build, test, and lint, python and install (editable, all extras).
Sponsio CLAUDE.md
Claude Code instructions for SponsioLabs/Sponsio, covering agent guide for sponsio, what sponsio is, positioning, what to read first and repository map.
agent-passport-system AGENTS.md
AGENTS.md instructions for aeoess/agent-passport-system, covering agents.md, dev environment, test before you ship, pr instructions and code style.
ProofFlow-v0.1 AGENTS.md
Instructions for Hyperion-GPU/ProofFlow-v0.1, covering proofflow agent rules, workflow, evidence, safety and architecture.
FrameCode-VibeWork AGENTS.md
AGENTS.md instructions for Sistema2D/FrameCode-VibeWork, covering agents.md, instruction order, reading-trigger rule, regression rule and when a plan is required.
agent-universe AGENTS.md
AGENTS.md instructions for SunrisesIllNeverSee/agent-universe, covering project state, original goal, architecture at a glance, what is built and functional and what is stubbed.