Skip to content
SkillModel-led3,014 tokens in use

C/C++ Security Review

Runs a multi-agent security review of native C and C++ userspace code and returns a merged report of memory safety, overflow and race findings.

8,319 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/c-review/skills/c-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 C/C++ Security Review costs

61tokens idle
3,014tokens 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 detectedwrites filesdetected

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 →
Good
  • Frontmatter12/15
  • Structure17/25
  • Examples12/15
  • Output15/15
  • Size13/15
  • 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 49
  • Writes files−8

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 50

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

C/C++ Security Review audits native userspace code for memory corruption, integer overflows, race conditions, type confusion and platform-specific bugs in Linux and macOS daemons or Windows services. It asks for four parameters, then makes one Workflow call that fans the work out to many review agents.

The workflow splits the code into review units, reviews slices in parallel, sweeps bug classes nobody touched, merges duplicates with a deterministic script, and writes REPORT.md, REPORT.sarif and findings.json. The skill states that no false-positive review ran, so the list is a set of leads and not adjudicated results.

What You Get

  • REPORT.md, REPORT.sarif and findings.json written to a timestamped .c-review-results folder
  • Findings with reviewer-assigned severity filtered to all, medium or high
  • A coverage figure comparing checks satisfied with checks required, with a warning when coverage is unmeasured
  • Clear flags when a run was partial, such as missing artifacts or a coverage gate that rejected the ledger

Common Use Cases

  • Auditing a native network daemon for remotely triggerable memory corruption
  • Reviewing a local privileged service under a local unprivileged attacker model
  • Scoping a review to one subdirectory of a large C library
  • Producing SARIF output to load into other security tooling

What Makes It Unique

Most C review prompts ask one model to read a repository. This skill partitions the code into units of at most 150 lines, assigns contiguous slices to separate agents, tracks a ledger of every unit and question, and gates the report on coverage, so partial runs are visible. It also refuses to present unadjudicated findings as confirmed.

Under The Hood

The skill asks you questions, then runs shell commands to create a results folder under the current directory and to resolve absolute paths. It requires uv, because a Python unit enumerator and a findings assembler run through it. It then makes one Workflow call that starts several sub-agents, which read your source files broadly for context and write part files, the report and a SARIF file into the results folder. It does not fetch web pages or call outside APIs, and it expects you to choose the model that the worker agents use.

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