Skip to content
SkillModel-led11,189 tokens in use

rust-review

A Rust security review run by parallel worker agents and two judges: safe/unsafe boundaries, memory safety, concurrency, panic DoS, FFI and async, reported as REPORT.md and SARIF.

8,320 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/rust-review/skills/rust-review
/plugin marketplace add trailofbits/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing rust-review costs

63tokens idle
11,189tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldeclared and detectednetworkdetectedwrites filesdeclared and detected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Excellent
  • Frontmatter12/15
  • Structure25/25
  • Examples12/15
  • Output15/15
  • Size6/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent
  • Runs shell commands−2

    does this on the machine, as the frontmatter declares

    SKILL.md, line 109
  • Writes files−4

    does this on the machine, as the frontmatter declares

    SKILL.md, line 154

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity100
  • Maintenance52
  • Community64
  • Maturity32
  • Momentum40

rust-review is a Trail of Bits skill for auditing Rust crates, services and libraries, especially those with unsafe code, FFI or concurrency. It detects which bug classes apply, plans clusters of checks with a script, and runs one worker agent per cluster in parallel.

A deduplication judge merges repeated findings, then a false-positive judge rules on each one and sets severity. The run ends with a REPORT.md and a SARIF file in an output directory.

What You Get

  • REPORT.md with the surviving findings and their severity
  • REPORT.sarif for tools that read SARIF, with an empty result list when nothing is found
  • Per-finding markdown files, coverage-gate files per worker, and run, dedup and false-positive summaries
  • A run summary that names any cluster that did not complete, so a partial run is never hidden

Common Use Cases

  • Auditing a Rust crate that contains unsafe blocks or an FFI layer
  • Looking for panic-induced denial of service in a Rust network service
  • Reviewing async runtime and concurrency code for deadlocks and atomic ordering mistakes

What Makes It Unique

A script, not the model, picks the clusters and renders the worker prompts, so the plan is the same on every run. Findings pass through a dedup judge and then a false-positive judge, and the skill states that the compiler’s borrow checker proves nothing about unsafe blocks, panics or FFI. It points secret-zeroization questions to zeroize-audit and C or C++ code to c-review.

Under The Hood

This is a large instruction file that drives many agents. The orchestrator runs shell commands: it checks for uv, searches the source tree with grep, rg and find, and creates an output directory. It runs the plugin’s Python scripts through uv to build the run plan, validate artifacts and generate SARIF, and it spawns named worker and judge subagents that read the code and write finding files to disk. It uses a question prompt to collect parameters. It does not fetch web pages or call outside APIs by its own instructions. Results are written under a results folder in the audited project by default.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers