audit-codebase

audit-codebase is a skill for Claude Code, Codex from runverdict/sf-security-review-toolkit. It costs 76 tokens per session (11,707 once invoked), scanned A, original, Apache-2.0.

A process for reviewing a codebase for security problems by having agents read and check the source code. It is a white-box review, meaning the reviewers can inspect the code, not a live attack against the running system.

In plain words
What is it for?
Use it to audit a partner's code, repeat a review after fixes, or check whether all instances of a reported security pattern were remediated.
Why use it?
It organizes security findings, checks them adversarially, and keeps a record so later reviews can focus on new or unresolved issues.

Skill for Claude CodeCodex

Installs and runs on its own, but its text points at files inside its plugin — anything it tells you to read at a ${CLAUDE_PLUGIN_ROOT} path is only there once the plugin is installed. Installing the plugin gets both.

Part of the sf-security-review-toolkit plugin — 14 skills, 1 hook 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/runverdict/sf-security-review-toolkit/audit-codebase
Any agent
npx skills add runverdict/sf-security-review-toolkit --skill audit-codebase
Clone the repo
git clone --depth 1 https://github.com/runverdict/sf-security-review-toolkit

Made for: Claude Code, Codex.

Or install sf-security-review-toolkit, the plugin that ships this one along with the rest of its 14 skills, 1 hook.

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-codebase

README.md
[![agentmods](https://agentmods.dev/badge/skills/runverdict/sf-security-review-toolkit/audit-codebase.svg)](https://agentmods.dev/skills/runverdict/sf-security-review-toolkit/audit-codebase)
Your own site
<a href="https://agentmods.dev/skills/runverdict/sf-security-review-toolkit/audit-codebase"><img src="https://agentmods.dev/badge/skills/runverdict/sf-security-review-toolkit/audit-codebase.svg" alt="Measured on agentmods" height="20"></a>
Per session 76 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 11,707 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 $0.00076 $0.11707
Opus 5 $0.00038 $0.05854
Sonnet 5 $0.00015 $0.02341
Haiku 4.5 $0.00008 $0.01171

Measured 4d ago against content hash 4fb60e480bc7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

audit-codebase 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.

skills/audit-codebase/SKILL.md · 660 lines

How it starts

The opening of the file, as written. The whole thing — 660 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Audit Codebase

Execute the audit engine specified in ${CLAUDE_PLUGIN_ROOT}/methodology/audit-methodology.md against the partner's repo. That spec is binding — pipeline, schemas, severity taxonomy, ledger mechanics, pass planning, and honesty constraints all live there; this skill is the operating procedure that runs it. The output is a verified findings report under docs/security-review/ plus an updated ledger under .security-review/. This is static code review by LLM agents reading source — never DAST, never a pen test (CONVENTIONS §2); Salesforce performs its own penetration testing regardless of anything this engine produces.

When to use

  • After /sf-security-review-toolkit:scope-submission wrote a scope manifest
  • Re-audit after remediation — cheap, because the ledger digest suppresses everything already found (the re-run loop, step 8)
  • After a failed review: ingest the failure report into the ledger as confirmed findings, then sweep the codebase for the same class — the remediation flow expects ALL instances of a flagged pattern fixed, not just the cited ones (baseline: process-failure-remediation-flow)
  • NOT for scanning packaged Apex — CRUD/FLS and the structured package pass belong to Code Analyzer in /sf-security-review-toolkit:run-scans
  • NOT a substitute for the authenticated DAST scan the review requires, and NOT for first-time architecture detection (/sf-security-review-toolkit:scope-submission)

Prerequisites

  • <target>/.security-review/scope-manifest.jsonrefuse to run without it (audit-methodology §1.1) and route to scope-submission. Dimension selection keys off the manifest; an audit of the wrong surface set is wasted spend
  • The dimension files in ${CLAUDE_PLUGIN_ROOT}/methodology/dimensions/
  • Baseline currency: read the baseline entries this skill leans on (process-failure-remediation-flow; the severity table in audit-methodology §4 points at process-review-fee and process-review-timeline) and warn if their last_verified is older than 90 days (CONVENTIONS §4)
  • A clean-enough working tree that file:line citations and ledger dedup keys stay meaningful across the run

Read the full file on GitHub · 660 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 · 660 lines · 76 tokens per session scan A 4fb60e480bc7

Subscribe to this mod's changes

audit-codebase is a skill published in the GitHub repository runverdict/sf-security-review-toolkit (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 76 tokens to every session and 11,707 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.

Related

Other skills, from other repositories