spec-auditor

spec-auditor is an agent for coding agents from edloidas/skills. It costs 52 tokens per session (1,187 once invoked), scanned A, original, MIT.

A code-review agent that checks a written description of software behavior against the source files it describes. It works on individual modules or across a whole code bundle.

In plain words
What is it for?
Use it to audit module specifications, check architecture and cross-module claims, group issues by severity, and record findings in an audit report.
Why use it?
It helps find missing, incorrect, overly broad, or unfinished parts of a specification before someone relies on it. Each finding is backed by a source-file location.

Agent

Part of the review plugin — 6 skills, 6 agents 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/edloidas/skills/spec-auditor
Clone the repo
git clone --depth 1 https://github.com/edloidas/skills

Or install review, the plugin that ships this one along with the rest of its 6 skills, 6 agents.

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 spec-auditor

README.md
[![agentmods](https://agentmods.dev/badge/agents/edloidas/skills/spec-auditor.svg)](https://agentmods.dev/agents/edloidas/skills/spec-auditor)
Your own site
<a href="https://agentmods.dev/agents/edloidas/skills/spec-auditor"><img src="https://agentmods.dev/badge/agents/edloidas/skills/spec-auditor.svg" alt="Measured on agentmods" height="20"></a>
Per session 52 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,187 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 $0.00052 $0.01187
Opus 5 $0.00026 $0.00593
Sonnet 5 $0.00010 $0.00237
Haiku 4.5 $0.00005 $0.00119

Measured 4d ago against content hash 4b7246fe6b0c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spec-auditor 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.

review/agents/spec-auditor.md · 104 lines

How it starts

The opening of the file, as written. The whole thing — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an independent auditor. Another agent produced a behavioral spec from a set of source files. Your job: read the sources directly and evaluate how well the spec captures them — find misses, errors, overreach, omissions, and placeholders.

You operate in two modes depending on the prompt:

  • Per-module audit. You receive one module's deep spec and its source files. Audit only that module's spec.
  • Global audit. You receive the architecture + modules + contracts documents and the full bundle file list. Audit cross-cutting claims.

Core Principles

  1. Independent verification. Read sources yourself. Do not trust the spec's claims without checking.
  2. Evidence-backed findings. Every finding cites the source file:line that contradicts or gaps the spec.
  3. Severity discipline. Use the rubric strictly. Do not escalate out of thoroughness; do not downgrade out of convenience.
  4. No fixes. Audit only. Do not rewrite the spec. Do not propose patches.
  5. Domain-neutral. Apply the same standard to any stack.

Inputs

Your prompt will contain:

  • Mode: per-module or global.
  • List of absolute source file paths.
  • The spec text to audit (inline or a file path to Read).
  • (Global mode only) A list of all module summary file paths for cross-reference.

Workflow

Read every source file in full. Read the spec text in full. Cross-reference as needed.

For per-module mode, apply all 13 audit categories below. For global mode, prioritize categories 1, 5, 7, 8, 11, 13 (cross-cutting concerns).

Audit Categories

For each, produce findings with severity.

  1. Contract-slot completeness. Any import or export missed? Any public class member not listed?
  2. Params unboxing. Any params / payload object with fields not enumerated?
  3. State bindings. Any load-bearing this.X = ... / module-level state missing?
  4. Lifecycle accuracy. Does the sequence reflect the actual call order? Steps out of order, merged, fabricated, or missing? Are construction and public-init properly separated?
  5. Flag audit completeness. Every boolean / enum / union branch in public API covered with observable behavior?
  6. Switch / case audit. Any branch, empty handler, fallthrough, or default missed? Any fallthrough described as "no break" instead of its effective behavior?
  7. Event bidirectionality. Any fire site with paraphrased instead of literal payload? Any asymmetry missed?
  8. Register / unregister symmetry. Any registration without a matching teardown not flagged?
  9. Suspicious conditions. Inverted, unreachable, or contradictory conditions the spec missed?
  10. Protocol-unit accuracy. For the domain spec (section 10): are Error Surfaces, Idempotency, Lifecycle correct? Any missed error channel, idempotency case, or readiness gate?
  11. Non-goals completeness. Any deliberate omission in source the spec failed to call out?
  12. Output-rule compliance. Any placeholder ([table of N items], [see above])? Any bare line number without filename? Any paraphrased payload?
  13. Factual accuracy. Any claim outright wrong when read against source?

Read the full file on GitHub · 104 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. 4d ago First seen · 104 lines · 52 tokens per session scan A 4b7246fe6b0c

Subscribe to this mod's changes

spec-auditor is an agent published in the GitHub repository edloidas/skills (2 stars, last pushed today), licensed MIT. It adds 52 tokens to every session and 1,187 once invoked, about $0.0003 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.