Skip to content
SkillModel-led380 tokens in use

replay-regression

Independently replay a regression proof of concept against freshly built before and after systems, and seal a record of whether the results agree.

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

Install command as written in the repository's README

Skill label

measured from its files

What installing replay-regression costs

21tokens idle
380tokens 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

Nothing detected: no shell, network, file writes, git or credentials in its files.

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
  • Structure16/25
  • Examples3/15
  • Output15/15
  • Size11/15
  • 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

replay-regression is the verification stage of the regression-hunt plugin from Trail of Bits. It takes a candidate, its frozen source records and a sealed reproduction package, then runs the same unchanged PoC in a fresh context against its own builds of the before (S0) and after (S1) systems.

The skill records whether S0 behaves correctly and S1 shows the claimed failure, checks that the interface is reachable with only the claimed user’s access, and writes a sealed replay.json. Completion means the replay finished, not that the regression is confirmed.

What You Get

  • A sealed replay.json that references the reproduction record and the original PoC and input
  • Separate logs, versions, commands and observations for the S0 and S1 runs
  • A list of every build directory used, kept apart from other attempts
  • A short report back: status, record path, one-sentence result and any blocker

Common Use Cases

  • Confirming that a regression PoC written by another agent works on a clean build
  • Checking that a claimed vulnerability is reachable through a real interface with the attacker’s own access
  • Recording a divergence between the reproducer’s logs and a second run before a report is written

What Makes It Unique

Replay is separated from reproduction on purpose. The replayer may not repair the PoC; a changed PoC goes back for new reproduction and replay. The skill also rejects trusting earlier logs and requires its own execution, which is a stricter check than most verification prompts ask for.

Under The Hood

The skill is a short prose checklist with no scripts of its own. It tells the agent to read the regression standard and evidence file contract in the plugin, then build and run the target systems, execute the PoC, save logs and write and seal a JSON record. That means running build and deploy commands and writing files inside an attempt directory, and it may start services the target needs.

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