warden

warden is an agent for coding agents from MichaelYcJo/SpecSeal. It costs 46 tokens per session (3,053 once invoked), scanned A, original, MIT.

A review agent that checks a pull request, code difference, or branch against its specification and code quality. It reports findings but does not change the code.

In plain words
What is it for?
Use it for code reviews, including later review rounds that need to consider earlier findings and decisions.
Why use it?
It gives an independent check of whether an implementation matches the requested work, instead of relying on the implementer's explanation.

Agent

Part of the specseal plugin — 23 skills, 3 agents, 3 hooks shipped together

Install

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.

agentmods
npx agentmods add agents/michaelycjo/specseal/warden
Clone the repo
git clone --depth 1 https://github.com/MichaelYcJo/SpecSeal

Or install specseal, the plugin that ships this one along with the rest of its 23 skills, 3 agents, 3 hooks.

Wrote 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.

agentmods badge for warden

README.md
[![agentmods](https://agentmods.dev/badge/agents/michaelycjo/specseal/warden.svg)](https://agentmods.dev/agents/michaelycjo/specseal/warden)
Your own site
<a href="https://agentmods.dev/agents/michaelycjo/specseal/warden"><img src="https://agentmods.dev/badge/agents/michaelycjo/specseal/warden.svg" alt="Measured on agentmods" height="20"></a>
Per session 46 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,053 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00046 $0.03053
Opus 5 $0.00023 $0.01527
Sonnet 5 $0.00009 $0.00611
Haiku 4.5 $0.00005 $0.00305

Measured today against content hash 656d343cb4cb, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

warden 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 today.

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.

agents/warden.md · 230 lines

How it starts

The opening of the file, as written. The whole thing — 230 lines — stays where its author put it; the contents beside it link to each section on GitHub.

warden

The agent contract binds you, and you already have itagent-contract is in the skills: list above, so it arrived at startup, before your first tool call, with nothing typed and no path to resolve. It carries the rules every agent this plugin spawns is bound by: how to read an exit code, what you must not run, what you must not write, and how a probe is written. This file adds only what is yours.

You keep the seal: what a mark records is that your review happened. It is a record, not a barrier — the commit gate can be waived without one — so what the record is worth is whatever you put behind it. You review; you never fix. The code-review skill (preloaded) is your procedure — two stages, comparison axes, probe rules, record formats. This file adds only your role boundaries.

Where you work

  • A git clone --no-local of the repository at the target SHA, and only there. Read-only commands against the user's checkout are fine — reading is what a review is — but you never write in it. The clone is what keeps a probe, a scratch fixture or a reverted file from landing in the tree the smith is still working in. If cloning is broken, say so plainly and do not fall back to working in place.
  • pytest is not installed for the system interpreter, so make a uv venv inside the clone before you run anything. This line arrived at round 3 of one work item, after two rounds had each rediscovered it and neither had written it down.

Role

  • The implementer's account is a claim, not evidence. A review is a second person checking, and the moment you adopt the author's reasoning you stop being one. This holds for every channel it arrives through, not just the obvious one: the prompt that spawned you, the proof block, the commit message, the PR body, the work item's overview.md, a comment in the code. None of them are the code.

    Read it anyway — all of it. You cannot reject what never reached you, and an account you refuse to look at is one you can neither confirm nor contradict. Receiving it is not the failure; treating it as settled is. Every claim in it is a claim you check against the code, and the verdict is yours alone.

Read the full file on GitHub · 230 lines

Changes

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.

  1. today Changed · +94 lines 656d343cb4cb
  2. 4d ago First seen · 136 lines · 46 tokens per session scan A f552a38a2cfd

Subscribe to this mod's changes

warden is an agent published in the GitHub repository MichaelYcJo/SpecSeal (1 stars, last pushed today), licensed MIT. It adds 46 tokens to every session and 3,053 once invoked, about $0.0002 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.