Skip to content
SkillModel-led1,302 tokens in use

fp-check

Verifies a suspected security bug and returns a TRUE POSITIVE or FALSE POSITIVE verdict backed by documented evidence.

8,319 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/fp-check/skills/fp-check
/plugin marketplace add trailofbits/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing fp-check costs

65tokens idle
1,302tokens 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

shelldeclaredwrites filesdeclared

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
  • Structure25/25
  • Examples0/15
  • Output10/15
  • Size10/15
  • Resources6/15

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

No findings.

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

FP Check takes a specific suspected vulnerability and decides whether it is real. It restates the claim, traces the data flow from source to sink, applies class-specific checks, and ends each bug with a verdict and its evidence.

Simple claims follow a linear checklist. Complex ones (cross-component paths, races, ambiguous logic bugs) go through a task-based deep verification with six mandatory gate reviews. It is for verifying findings, not for hunting new ones.

What You Get

  • A TRUE POSITIVE or FALSE POSITIVE verdict for every suspected bug
  • Documented evidence: data flow traces, attacker control analysis and devil’s advocate reviews
  • A final summary with counts and a short reason for each rejection
  • Class-specific checks for memory corruption, logic bugs, races, integer issues, crypto, injection, disclosure, denial of service and deserialization

Common Use Cases

  • Checking whether a static analyzer finding is exploitable
  • Validating a bug report from a bounty submission
  • Triaging a batch of scanner results before they reach developers
  • Confirming a suspected race condition or logic flaw

What Makes It Unique

It is built to counter the bias of language models toward seeing bugs and overrating severity. A table of rationalizations to reject blocks shortcuts such as reporting code that merely looks unsafe, and each bug must pass all mandatory gate reviews. It also keeps discovery out of scope, so it verifies claims and does not search for new ones.

Under The Hood

The skill is prose with reference files for standard verification, deep verification, gate reviews, bug classes, false positive patterns and evidence templates. The agent reads and searches the code under review, may run shell commands to test a claim, can write evidence files, and in deep mode dispatches the plugin’s analysis agents and tracks phases as tasks. It asks you questions when a claim is unclear. It makes no web requests of its own.

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