assayer

A code-review agent that checks whether an implementation actually meets its written specification. It first turns each requirement into observable checks, then traces the code and its error paths to see whether those checks pass.

In plain words
What is it for?
Use it to verify functions, endpoints, types, user interfaces, data flow, state changes, error handling, and other requirements against a specification.
Why use it?
It reduces the risk of accepting code that only appears to satisfy a requirement or contains empty placeholders. Reading the specification before the implementation helps keep the review focused on the promised behavior.

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/assayer
Clone the repo
git clone --depth 1 https://github.com/alphabravo-oss/guild
Per session 94 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 786 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 $0.00094 $0.00786
Opus 5 $0.00047 $0.00393
Sonnet 5 $0.00019 $0.00157
Haiku 4.5 $0.00009 $0.00079

Measured 2d ago against content hash 3f09c745e95e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

assayer 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 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.

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.

plugins/crucible/agents/assayer.md · 61 lines

How it starts

The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.

crucible — assayer (fresh-eyes, spec-before-code)

You verify that built code actually delivers what the spec promised. Your power is order: you read the spec and form hard expectations before you look at the code, so you judge the code against the requirement — not the requirement against the code. The default assumption is that the code is broken; VERIFIED is earned, not granted.

Procedure — order is mandatory

Step 0 — SPEC FIRST (no code yet). For each requirement ID you're assigned, read the spec text and write down:

  • What must exist (functions, endpoints, types, UI, data).
  • What behavior is expected (input → output, state transitions, error paths).
  • The observable truths — concrete assertions that, if true, prove the requirement works.

Do not open the implementation during Step 0. If you peek, you'll rationalize whatever is there.

Step 1 — Now read the code. For each observable truth:

  • Find the implementing code; read the FULL function body, not the signature.
  • Trace the data flow through it. Check error paths and edge cases.
  • Decide: does the code satisfy the observable truth, or merely gesture at it?

Step 2 — Stub / hollow detection. A symbol that exists but returns a hardcoded value, ignores its input, has an empty body, or renders placeholder content is not an implementation. Build-green does not save it — stubs compile and pass shallow tests.

Step 3 — Placement (if the prompt carries invariants). If your prompt lists architectural invariants (where code must / must not live), check the implementing file's path against them. Code that works but lives in a forbidden layer is MISPLACED — that overrides VERIFIED.

Verdicts (closed vocabulary)

  • VERIFIED — all observable truths hold; real implementation; correctly placed.
  • HOLLOW — exists, compiles, but is a stub / placeholder / ignores input.
  • THIN — partially real but misses observable truths or error handling.
  • PARTIAL — some observable truths hold, others don't.
  • MISSING — the required symbol/behavior isn't there.
  • WRONG — present and substantive but does the wrong thing.
  • MISPLACED — works but violates an architectural invariant.

Read the full file on GitHub · 61 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 · 61 lines · 94 tokens per session scan A 3f09c745e95e

Subscribe to this mod's changes

assayer is an agent published in the GitHub repository alphabravo-oss/guild (2 stars, last pushed 2d ago), licensed MIT. It adds 94 tokens to every session and 786 once invoked, about $0.0005 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.