Skip to content
SkillModel-led1,651 tokens in use

patching-bugs

The orchestrator for fixing one known bug: reproduce it, validate it, patch it, prove the patch, hunt regressions, and render reports for review.

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

Install command as written in the repository's README

Skill label

measured from its files

What installing patching-bugs costs

30tokens idle
1,651tokens 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

shelldetected

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
  • Examples9/15
  • Output15/15
  • Size10/15
  • 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

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

Patching-bugs coordinates the full repair of one named bug in a local Git repository. It normalizes the report into a finding record, then hands each stage to a fresh agent: reproduction, independent validation, up to three patch attempts, independent proof, and a regression hunt over the patch range.

Every handoff is a validated JSON record with a hash manifest, and results are transcribed from receipts. The run ends with an auditor report and, only when the patch is verified, a maintainer report. It never pushes or opens a pull request.

What You Get

  • A run directory outside the target tree holding the finding, reproduction, validation, patch, proof and regression records
  • One local patch commit with fix.patch and tests.patch for each attempt
  • A report.html for the auditor on every run, including failed and inconclusive ones
  • A maintainer-report.html only when the validated records establish a verified patch
  • A verdict such as ready, ready with concerns, not ready or inconclusive, derived from receipts

Common Use Cases

  • Fixing a security bug an auditor reported against a local codebase
  • Checking a compile failure or runtime crash against its stated invariant before changing code
  • Producing reviewable proof that a patch fixes the bug and breaks nothing nearby

What Makes It Unique

Different fresh agents do reproduction, validation, patching and proof, and the validator never sees the reproducer’s transcript. A runtime reproduction counts only when the exact PATCH_REACHED marker appears. The skill lists shortcuts to reject, such as trusting the patcher’s own test run or treating a missing receipt as close enough.

Under The Hood

The agent reads SKILL.md, a record contract and JSON schemas in the plugin, and runs bundled Python scripts through uv to validate records, write manifests, run a three-arm proof and render reports. It runs the project’s own build and test commands, uses Git to read baseline objects, create worktrees and make one local commit, dispatches other agents, and writes many files into a run directory. It does not push or open pull requests, and the supplied bug report is treated as data.

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