Skip to main content
skill CX Codex

zeroize-audit — Claude Code skill: audit secret wiping

Finds passwords, keys and tokens left in memory across C, C++ and Rust, and checks whether the compiler threw their wiping away.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add trailofbits/skills --skill zeroize-audit -a claude-code
Open on GitHub

What zeroize-audit does

The skill walks the source tree looking for sensitive values that never get cleared after use. It also checks whether the compiler removed the wipe itself, and it only reports an optimized-away finding when there is evidence from IR or assembly. To get that evidence it compares builds at several optimization levels, O0, O1 and O2 by default. O1 is the diagnostic one: a wipe that vanishes there is plain dead-store elimination, while O2 catches more aggressive cases.

Execution runs through phases 0 to 7, each driven by its own workflow file, with the author's prompt defining the orchestrator role. There is control-flow verification, tracking of secret copies, heap allocator warnings, and generation of runtime tests. Analysis is read-only against your codebase; artifacts go to a temporary working directory. The output is a JSON report.

You need the repo root plus at least one of compile_commands.json for C and C++ or Cargo.toml for Rust. Preflight stops the run without one, so have the compile database path ready. The skill also leans on the serena MCP server: no symbol or reference lookup without it. An interrupted run can resume from its saved phase when the working directory is known.

What it helps with

  • find keys and passwords still sitting in memory after use
  • check whether the compiler dropped a secret wipe
  • back an optimized-away finding with IR or assembly evidence
  • trace where copies of secret values end up
  • generate a runtime test that catches the leak

Example prompts for the agent

  • Audit this crate for secrets left in memory
  • Check that this C library actually wipes its keys
  • Run zeroize-audit against compile_commands.json and show me the report

How to install zeroize-audit

1Skills CLI (Vercel)

npx skills add trailofbits/skills --skill zeroize-audit -a claude-code

2Manually: put the files in place

~/.claude/skills/zeroize-audit/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/zeroize-audit/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about zeroize-audit

How do I install zeroize-audit in Claude Code?

Run npx skills add trailofbits/skills --skill zeroize-audit -a claude-code. The skill lands in ~/.claude/skills/zeroize-audit/SKILL.md.

What does zeroize-audit do?

It detects missing zeroization of sensitive data in C, C++ and Rust and checks whether compiler optimizations removed it. An optimized-away finding requires IR or assembly evidence.

What do I need before running it?

The repo root and a path to compile_commands.json for C and C++ or to Cargo.toml for Rust. Preflight halts the run without one of them.

Similar skills

Everything for Security →

From the same set77

ruzzy This skill installs and runs Ruzzy, a coverage-guided Ruby fuzzer built on libFuzzer. Reach for it when a gem needs crash hunting or a native extension needs a memory-safety check. CX Codex +1 5,100installs wycheproof Validates cryptographic implementations using Project Wycheproof test vectors. It helps your agent spot known vulnerabilities and edge cases across AES, RSA, and ECDSA. CX Codex +1 5,200installs skill-improver This skill runs repeated review and fix rounds on another skill until critical and major findings are gone. CX Codex +1 5,000installs let-fate-decide This skill uses a Tarot spread to suggest a direction when a request leaves several reasonable options open. It suits casual handoffs where the agent is asked to choose. CX Codex +1 4,700installs diagramming-code Generates Mermaid diagrams from code graphs using Trailmark. Useful for visualizing call paths, module dependencies, and data flows. CX Codex +1 4,200installs dimensional-analysis Audits numeric code for unit confusion and decimal-scale errors, then adds dimensional comments. Use it when formulas or conversions need a unit-level review. CX Codex +1 4,200installs mutation-testing A skill that walks the agent through mutation testing: setting up a mewt or muton campaign, reading surviving mutants, hunting bugs in untested code. CX Codex +1 4,200installs trailmark Builds and inspects multi-language code graphs to trace taint propagation, map attack surfaces, and prioritize security audits. CX Codex +1 4,000installs

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor