soundcheck CLAUDE.md

soundcheck CLAUDE.md is an instructions file for coding agents from thejefflarson/soundcheck. It costs 2,821 tokens per session, scanned A, original, MIT.

Project instructions for Soundcheck, a Claude Code plugin that reviews code for security problems through automatic checks and on-demand scans.

In plain words
What is it for?
They are for maintaining Soundcheck, running pull-request or full-repository security reviews, and authoring its skills.
Why use it?
They explain the project’s development conventions, review modes, threat model, and rules for adding security skills.

Instructions file

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 instructions/thejefflarson/soundcheck/claude-md
Clone the repo
git clone --depth 1 https://github.com/thejefflarson/soundcheck

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 soundcheck CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/thejefflarson/soundcheck/claude-md.svg)](https://agentmods.dev/instructions/thejefflarson/soundcheck/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/thejefflarson/soundcheck/claude-md"><img src="https://agentmods.dev/badge/instructions/thejefflarson/soundcheck/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 2,821 This file is loaded in full into every session.
When invoked 2,821 The same file — it is already loaded in full.
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 $0.02821 $0.02821
Opus 5 $0.01411 $0.01411
Sonnet 5 $0.00564 $0.00564
Haiku 4.5 $0.00282 $0.00282

Measured yesterday against content hash 07b364721599, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

soundcheck CLAUDE.md 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 yesterday.

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.

CLAUDE.md · 240 lines

How it starts

The opening of the file, as written. The whole thing — 240 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Soundcheck — Dev Conventions

What is Soundcheck?

Soundcheck is a Claude Code plugin providing 48 auto-invoking security skills plus three on-demand review modes (and one cleanup orchestrator), each scoped to a distinct use case. When Claude detects vulnerable code patterns mid-task, the relevant per-category skill auto-invokes, flags the issue, explains the change, and continues — no user intervention required.

Auto-invocation is driven entirely by the description frontmatter in each SKILL.md. No CLAUDE.md trigger mapping is needed — the description field IS the trigger.

Three review modes

The on-demand reviews are split into three discrete modes, each with its own skill, driver script, latency profile, and target bug class. Reach for the one that matches the situation; don't try to make one mode do another's job.

Mode Skill Driver Latency Model Target
1 — PR gate pr-review security-review-action.py --diff-base REF ≤1 min haiku Critical/High OWASP in changed files
2 — Full scan security-review security-review-action.py --full-repo ~20 min sonnet/opus All severities, whole repo, with subagent fan-out
3 — Contract review contract-review contract-review.py --rounds N ~30 min opus Caller/callee invariant gaps that single-pass review misses

Mode 2 dispatches 5 subagents (threat-modeling, hotspot-mapping, design-review, vulnerability-audit, attack-chain-analysis). Mode 1 is single-pass — no subagent dispatch — and filters to Critical/High only so the gate stays fast and quiet. Mode 3 dispatches threat-modeling, hotspot-mapping, and one contract-audit per round. Output is the findings table printed to stdout — no on-disk state. A Mythos-style trust-anchor-confusion bug that mode 2 missed twice (Botan certificate_known) is the canonical mode 3 target.

Threat model (for /security-review against this repo)

  • Purpose: Claude Code plugin (skills) + a PR-gate GitHub Action (scripts/security-review-action.py) + a GitHub Action-style self-test.
  • Attack surface: scripts/security-review-action.py (CI with write perms and API access), .github/workflows/*, .claude-plugin/plugin.json, .claude/skills/*/SKILL.md. Everything under scripts/ is maintainer code run locally or in CI with trusted inputs.
  • Trusted inputs: maintainer-committed skill files, scripts/, docs/skill-template.md, plugin.json, hardcoded benchmark manifests.
  • Untrusted inputs: third-party repo content scanned by security-review-action.py, cloned benchmark repos, GitHub issue content.
  • Out of scope (auditors should discount these):
    • Defense-in-depth hardening on local-only dev scripts (benchmark-*.py, smoke-test-skills.py, validate-skills.py).
    • Intentionally vulnerable fixtures in docs/test-cases/.
    • Opt-in flags whose behavior is the feature: --apply-rewrites on security-review-action.py hands the agent Edit access by design so CI can apply fixes automatically; this is opt-in, not a default.
    • Word-count / lint / style issues in skill files.

Read the full file on GitHub · 240 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. yesterday First seen · 240 lines · 2,821 tokens per session scan A 07b364721599

Subscribe to this mod's changes

soundcheck CLAUDE.md is an instructions file published in the GitHub repository thejefflarson/soundcheck (20 stars, last pushed 1mo ago), licensed MIT. It adds 2,821 tokens to every session, about $0.0141 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 2026-09-03.