zeroize-audit

An agent skill by trailofbits, from trailofbits/skills. Tags: assembly, code-review, formal-verification, rust, security.

What it does

Detects missing zeroization of sensitive data in source code and identifies zeroization removed by compiler optimizations, with assembly-level analysis, and control-flow verification. Use for auditing C/C++/Rust code handling secrets, keys, passwords, or other sensitive data.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add trailofbits/skills --skill zeroize-audit

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/trailofbits/skills
cp -r skills/plugins/zeroize-audit/skills/zeroize-audit ~/.claude/skills/zeroize-audit

Safety box score

Grade D, elevated risk: 40 of 100 risk points, rated 2026-09-22. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is an analysis pipeline, but it still compiles, runs, and in some phases executes generated PoCs against the target repo. The main risks are broad code execution through build tooling, network fetches for Serena via `uvx`, and hidden behavior in several unreadable helper scripts.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	3 (high)	It explicitly runs build and analysis commands such as `cargo +nightly check`, `cargo rustc`, `clang`, and Phase 5 says to `compile and run all PoCs`, so it can execute repository build scripts and generated programs.
Network access	2 (moderate)	It fetches Serena from GitHub with `uvx --from git+https://github.com/oraios/serena`, and Rust/Cargo analysis may also resolve dependencies over the network.
Filesystem reach	1 (low)	Writes analysis artifacts only into `/tmp/zeroize-audit-{run_id}/` and cleans temp dirs with `rm -rf` traps; there is no evidence of broad writes outside the workdir.
Secrets & credentials	1 (low)	The skill is designed to inspect code that handles `key`, `token`, `password`, and similar data, so it may read sensitive source material, but it does not access keychains, shell histories, or other secret stores.
Privilege & persistence	0 (none)	The skill itself does not request sudo, global installs, service registration, or shell/OS setting changes; the only install references are external prerequisites like `apt install bear`.
Instruction hijack surface	0 (none)	I do not see instructions to ignore higher-priority policies or hide actions; the workflow text is imperative, but that is normal orchestration rather than prompt injection.
Irreversible actions	1 (low)	It automatically generates, validates, and optionally tests PoCs (`Phase 4`/`Phase 5`/`Phase 7`) without a human check, but it does not deploy, publish, message, pay, or book anything.
Transparency	3 (high)	Several core helpers could not be read, including `tools/generate_poc.py`, `tools/scripts/check_llvm_patterns.py`, `tools/scripts/check_mir_patterns.py`, `tools/scripts/check_rust_asm.py`, `tools/scripts/semantic_audit.py`, and `tools/scripts/find_dangerous_apis.py`.

Before installing

Inspect `references/mcp-analysis.md` and decide whether you want a skill that can run `uvx --from git+https://github.com/oraios/serena`.
Run the workflow in a disposable or non-production checkout first, because it may execute build scripts and generated PoCs.
Review `tools/generate_poc.py` and the other unreadable helper scripts in a trusted environment before enabling the skill.
If network egress matters, block it and confirm the skill still works without fetching MCP tooling or Cargo dependencies.

Rated 2026-09-22 by an automated review of 38 files from the skill's repository.

Source

Repository: trailofbits/skills (all skills from this repository)
Path: plugins/zeroize-audit/skills/zeroize-audit/SKILL.md
Branch: main
Collection: zeroize-audit
Updated: 2026-09-19

trailmark-review-gate — Runs a Trailmark structural review gate over a branch, pull request, fix commit, release diff, or git ref range to detect new entrypoints, new tainted paths.
diff-scanning-with-aws-security-agent — Run a fast AWS Security Agent diff scan on only the changed code since a git ref.
scanning-with-aws-security-agent — Run an AWS Security Agent scan on the workspace — uploads the source to AWS, scans it with the managed Security Agent service, and returns ranked.
thermo-nuclear-review — Comprehensive security and correctness audit of a branch's changes.
thermos — Launch both thermo-nuclear review subagents in parallel, then synthesize their findings.
address-sanitizer — Builds and runs code under AddressSanitizer to catch buffer overflows, use-after-free, and other memory errors during fuzzing or tests.