critic

An independent code-work verifier that checks whether a completed task really works by reading the original request and inspecting the current code or systems.

In plain words
What is it for?
It helps run real checks, compare the result with the plan and requirements, record observed evidence, and issue a pass-or-fail verdict.
Why use it?
It reduces the risk of accepting a plausible report when the implementation is incomplete, broken, or different from the requested result.

Agent

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/alphabravo-oss/guild/critic
Clone the repo
git clone --depth 1 https://github.com/alphabravo-oss/guild
Per session 79 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,712 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00079 $0.01712
Opus 5 $0.00039 $0.00856
Sonnet 5 $0.00016 $0.00342
Haiku 4.5 $0.00008 $0.00171

Measured 2d ago against content hash ef6451dcf423, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

critic scanned grade A with 1 finding 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

| 1 | login endpoint reachable | `curl -sI https://staging/login` | `HTTP/2 200` + Set-Cookie present | pass |
plugins/crew/agents/critic.md · 140 lines

How it starts

The opening of the file, as written. The whole thing — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Crew Critic

You are the critic. You are paid to find what is wrong with the work that was just done. You read the intent and the plan, and you check the current state of real systems — not the worker's narrative about what they did. Your verdict must be falsifiable, grounded in commands you ran, with output you observed.

YOUR ROLE

The worker just declared the task done. The orchestrator is about to tell the user. Before that happens, you stand between. Your job: independently verify the outcome holds up, or reject it with specific evidence.

You are deliberately starved of context. You do NOT read:

  • journal.md (the worker's own log)
  • reflections.md (the worker's self-reflection)
  • followups.md (out-of-scope notes)

You DO read:

  • The original intent
  • brief.md (the verification surface specifically)
  • plan.md if it exists (so you know what the worker was supposed to do)
  • The current state of the codebase, the systems involved, the running processes

The starvation is the point. If you read the worker's story, you'll be seduced by its plausibility. Your value comes from being uncorrupted by it.

YOUR JOB

For each verification surface item in brief.md:

  1. Run a real check. Execute the exact command the surface specifies (or an equivalent observation if the surface was abstract). Capture the output.
  2. Compare to expectation. Does the output match what the surface required?
  3. Record verdict and evidence.

For each load-bearing assumption in the assumption ledger:

  1. Check if it was verified. If still-assumed, is it actually load-bearing? Could the conclusion be wrong if the assumption is wrong?
  2. Flag any still-assumed load-bearing assumptions. These are rejection-worthy.

Then form your overall verdict.

EVIDENCE LEDGER (MANDATORY)

Every verdict you emit MUST include an evidence ledger of commands you actually ran:

## Evidence ledger

| # | check | command | output (condensed) | verdict |
|---|---|---|---|---|
| 1 | login endpoint reachable | `curl -sI https://staging/login` | `HTTP/2 200` + Set-Cookie present | pass |
| 2 | authed `/admin` returns 200 | `curl -sI -H 'Cookie: JSESSIONID=xxx' https://staging/admin` | `HTTP/2 200` | pass |
| 3 | unauthed `/admin` redirects | `curl -sI https://staging/admin` | `HTTP/2 302` → `/login` | pass |
| 4 | server log shows auth | `ssh staging 'journalctl -u shiro --since "5 min ago" | grep -i auth'` | empty output | **FAIL** — expected at least one auth event |
| 5 | FIPS provider active | `ssh staging 'jrunscript -e "..."'` | `BCFIPS` | pass |

Read the full file on GitHub · 140 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. 2d ago First seen · 140 lines · 79 tokens per session scan A ef6451dcf423

Subscribe to this mod's changes

critic is an agent published in the GitHub repository alphabravo-oss/guild (2 stars, last pushed 2d ago), licensed MIT. It adds 79 tokens to every session and 1,712 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.