Use when creating, adapting, reviewing, or debugging JS/TS fuzzers, corpus fixtures, replay files, shrinkers, repro writers, or deterministic random test harnesses.
Installs into .claude/skills of the current project.
Are you the author of Fuzz Harness Builder?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/siliconjungle-fuzz-harness-builder)
---
name: fuzz-harness-builder
description: Use when creating, adapting, reviewing, or debugging JS/TS fuzzers, corpus fixtures, replay files, shrinkers, repro writers, or deterministic random test harnesses.
---
# Fuzz Harness Builder
Fuzzers should produce useful failures, not just random load.
## Required Properties
- Deterministic seed: `--seed`.
- Bounded run size: `--cases`, and target-specific caps where needed.
- Replayable failure: `--replay` or `--write-repro`.
- Stable oracle: compare against a target-owned contract from docs, tests, source, or fetched material.
- Corpus promotion path: failures become checked-in fixture cases.
- Source traceability: corpus ideas from external projects cite the source-pass note or fetch manifest.
## Workflow
1. Identify the contract from the target project.
2. Generate valid values first, then mutate them according to the target contract.
3. Add or adapt a template:
```sh
npx evidence-kit add-fuzzer --name core --language js
```
4. Replace `runSubject()` with the project API under test.
5. On failure, write a reduced JSON repro and add it to the corpus.
6. When fuzzing ideas come from external sources, run `npm run research:list` and cite the matching source note.
## Reporting
Final reports should include command, seed, cases, any replay path, and whether corpus fixtures were added.
