audit-correctness

audit-correctness is a skill for Claude Code from JHostalek/dotclaude. It costs 72 tokens per session (2,757 once invoked), scanned A, original, CC0-1.0.

A method for finding logic errors where code produces the wrong result or state, including boundary mistakes, incorrect conditions, and unit mismatches.

In plain words
What is it for?
Use it for correctness reviews and for fixing bugs such as off-by-one errors, inverted conditions, missing edge cases, or code that contradicts its documentation.
Why use it?
It checks the intended behaviour across the full path from input to stored data and later cleanup, including retries, failures, and older code paths.

Skill for Claude Code

Written for Claude Code: argument-hint in frontmatter.

Part of the jhostalek-skills plugin — 32 skills, 12 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 skills/jhostalek/dotclaude/audit-correctness
Any agent
npx skills add JHostalek/dotclaude --skill audit-correctness
Clone the repo
git clone --depth 1 https://github.com/JHostalek/dotclaude

Made for: Claude Code.

Or install jhostalek-skills, the plugin that ships this one along with the rest of its 32 skills, 12 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 audit-correctness

README.md
[![agentmods](https://agentmods.dev/badge/skills/jhostalek/dotclaude/audit-correctness.svg)](https://agentmods.dev/skills/jhostalek/dotclaude/audit-correctness)
Your own site
<a href="https://agentmods.dev/skills/jhostalek/dotclaude/audit-correctness"><img src="https://agentmods.dev/badge/skills/jhostalek/dotclaude/audit-correctness.svg" alt="Measured on agentmods" height="20"></a>
Per session 72 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,757 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 5.1 $0.00072 $0.02757
Opus 5 $0.00036 $0.01378
Sonnet 5 $0.00014 $0.00551
Haiku 4.5 $0.00007 $0.00276

Measured 5d ago against content hash 4ee1841472c6, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

audit-correctness 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 5d 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.

skills/audit-correctness/SKILL.md · 144 lines

How it starts

The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.

!cat "${CLAUDE_SKILL_DIR}/../shared/audit-workflow.md"

Run as the correctness dimension. Find behavior that can produce an observably wrong result or state, whether the defect is local, architectural, emergent between components, or visible only under particular lifecycle or operating conditions.

Work top-down

Start from intended system behavior, not grep patterns:

  1. Reconstruct the behavioral model from user-visible promises, domain rules, schemas, types, APIs, state machines, callers, tests, migrations, configuration, deployment topology, and operational workflows. Identify authoritative state, derived state, invariants, preconditions, postconditions, ownership, ordering, consistency guarantees, and allowed failure behavior.
  2. Map each material behavior across its complete path: input or event, normalization, validation, computation, state transition, persistence, replication or messaging, caching, output, and later reconciliation or cleanup. Include alternate, legacy, administrative, asynchronous, retry, recovery, and mixed-version paths.
  3. Derive consequences of divergence before inspecting implementation detail: wrong result, lost or duplicated work, invalid state, stale view, irreversible action, broken accounting, cross-component disagreement, or violation that appears only under load, restart, failover, clock change, or concurrent use.
  4. Inspect architecture-wide invariants and dangerous compositions first, then trace exact call paths and apply the baseline to every applicable component and boundary. Search for variants after confirming a flaw.

Treat names, signatures, docstrings, comments, caller expectations, tests, specifications, and existing behavior as evidence of intent, not independent truth. Resolve conflicts among them using product behavior and system invariants. If intent remains ambiguous, retain an unresolved question rather than selecting a convenient interpretation.

Never infer correctness from a framework, type system, schema, ORM, transaction helper, generated client, shared utility, naming convention, familiar pattern, or passing test. Verify what the exact version, configuration, call path, data shape, isolation level, deployment topology, and failure timing guarantee. Individually reasonable components can compose into incorrect behavior.

Read the full file on GitHub · 144 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. 5d ago First seen · 144 lines · 72 tokens per session scan A 4ee1841472c6

Subscribe to this mod's changes

audit-correctness is a skill published in the GitHub repository JHostalek/dotclaude (11 stars, last pushed yesterday), licensed CC0-1.0. It adds 72 tokens to every session and 2,757 once invoked, about $0.0004 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.