reproduce-regression — Claude Code skill for regression testing
The reproduce-regression Claude Code skill reproduces a regression candidate on live systems before and after a change. It builds a PoC and compares both versions using the same input.
CC-BY-SA-4.0 671 installs ★ 8,341 updated 2026-10-24
npx skills add trailofbits/skills --skill reproduce-regression -a claude-codeWhat reproduce-regression does
The skill deploys S0 and S1, traces the candidate to a real user or attacker entry point, and tests the reported regression. It follows the project guidance, regression standard, and evidence file requirements before working on the target.
It creates a portable PoC and runs it against both deployments with identical input and comparable configuration. The test uses live systems, not mocks, extracted functions, or simulations. reproduce.json records the setup, tool and dependency versions, commands, revisions, logs, observations, and false-positive checks.
Builds and deployments stay inside the attempt directory, and every build directory is listed in build_dirs. Temporary resources are released while evidence is kept in a sealed package. A finished experiment does not itself confirm a regression.
What it helps with
- compare system behavior before and after a change
- test a regression candidate through a real entry point
- create a repeatable PoC for two deployments
- record experiment conditions and check for false positives
Example prompts for the agent
- Reproduce this regression candidate on S0 and S1 and save the results.
- Build a PoC for this regression against both live deployments using identical input.
- Check whether the reported failure reproduces and record it in
reproduce.json.
How to install reproduce-regression
1Skills CLI (Vercel)
npx skills add trailofbits/skills --skill reproduce-regression -a claude-code2Manually: put the files in place
~/.claude/skills/reproduce-regression/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/reproduce-regression/
1Skills CLI (Vercel)
npx skills add trailofbits/skills --skill reproduce-regression -a codex2Manually: put the files in place
.agents/skills/reproduce-regression/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add trailofbits/skills --skill reproduce-regression -a cursor2Manually: put the files in place
.cursor/skills/reproduce-regression/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install trailofbits/skills plugins/regression-hunt/skills/reproduce-regression2Manually: put the files in place
.github/skills/reproduce-regression/SKILL.md1Skills CLI (Vercel)
npx skills add trailofbits/skills --skill reproduce-regression -a gemini-cli2Platform CLI
gemini skills install https://github.com/trailofbits/skills --path plugins/regression-hunt/skills/reproduce-regressionManually: put the files in place
.devin/skills/reproduce-regression/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add trailofbits/skills --skill reproduce-regression -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about reproduce-regression
How do I install reproduce-regression in Claude Code?
Run npx skills add trailofbits/skills --skill reproduce-regression -a claude-code. The skill file is placed at ~/.claude/skills/reproduce-regression/SKILL.md.
What does reproduce-regression do?
It runs a PoC on live systems from before and after a change, records their behavior, and packages the evidence. A completed experiment does not necessarily confirm the regression.
How does this differ from independently checking an existing PoC?
reproduce-regression builds a PoC and tests it against two live deployments. The separate replay workflow is for independent verification of an existing PoC.