Back to search
Witnis
SkillGitHub
Indexes your repo. On a PR it comments what the change touches, and whether tests actually ran.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/l00plegend--witnis)HTML
<a href="/listings/l00plegend--witnis"><img src="/api/listings/l00plegend--witnis/badge.svg" alt="findsafeskills safety rating" /></a>Install
i
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
pip install witnis # or uvx witnis
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 74 lines
- <!-- mcp-name: io.github.L00PLeGeNd/witnis -->
- # Witnis
- Local evidence for coding agents: index a repo, show blast radius, put the same envelope on a PR, and keep verify **blocked** until you authorize it.
- Search name: **witnis-dev**. Not [Witnis AB](https://www.witnis.se). Apache-2.0.
- ## Install
- ```bash
- pip install witnis
- # or
- uvx witnis
- ```
- ## What it catches that grep misses
- On [spring-petclinic](https://github.com/spring-projects/spring-petclinic) @ `500158f`, `git grep -lw Person -- '*.java'` hits five files. None of them are `OwnerController`, `PetController`, or `VisitController` — those only talk about `Owner`.
- `witnis impact Person` follows `Owner extends Person` into those controllers.
- ```bash
- witnis index .
- witnis impact Person --plain
- ```
- That is analysis completeness (grep vs index), not an LLM A/B test. Dual-human labels are still open. Notes: [`docs/benchmarks/petclinic-person-grep-vs-impact.md`](docs/benchmarks/petclinic-person-grep-vs-impact.md).
- <p align="center">
- <img src="./docs/assets/blast-radius.svg" alt="git grep Person finds five files; witnis impact Person reaches OwnerController, PetController, VisitController" width="960"/>
- </p>
- ## PR Action
- Use a full clone (`fetch-depth: 0`). `@v1` moves with 1.x; pin `@v1.11.2` if you do not want that. Default run is blocked — 0 commands executed.
- ```yaml
- - uses: actions/checkout@v4
- with: { fetch-depth: 0 }
- - uses: L00PLeGeNd/Witnis/.github/actions/pr-gate@v1
- ```
- ## MCP
- Public tools: `index`, `look`, `impact`, `verify`. Everything else needs `WITNIS_MCP_EXPERIMENTAL=1`.
- ```bash
- uvx witnis loop-init
- ```
- That writes a Cursor / Claude snippet and a Harnis stdio policy. One install — seatbelt is not a second product. `WITNIS_DB` defaults to `.witnis/index.sqlite`.
- ```json
- {
- "mcpServers": {
- "witnis": {
- "command": "npx",
- "args": [
- "-y", "harnis", "mcp", "--preset", "coding-agent", "--",
- "uvx", "witnis", "mcp"
- ]
- }
- }
- }
- ```
- ## Verify honesty
- Verify does nothing until `WITNIS_ALLOW_VERIFY=1` and `--execute`. `automation_verified` means the allowlisted commands ran. It is not a human review.
- More: [`SECURITY.md`](SECURITY.md) · [`docs/trusted-agent-loop.md`](docs/trusted-agent-loop.md) · [`docs/cli-cheatsheet.md`](docs/cli-cheatsheet.md) · PetClinic GT (not dual-human): [`docs/benchmarks/petclinic-human-v2.md`](docs/benchmarks/petclinic-human-v2.md).
- mcp-name: io.github.L00PLeGeNd/witnis
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.