reverse-skill
reverse-skill Deployment: Pin Clients, Tools, and Evidence Boundaries
Deploy reverse-skill safely across Windows, Ubuntu, Kali, and compatible AI clients with pinned tools, isolated cases, and regression gates.
What you will learn
- Plan a cross-platform reverse-skill deployment
- Operate routing CI and supply-chain gates
- Design recovery and retention controls
Before you start
- Basic Git and command-line usage
- Comfort reading a project README
You can explain the project, run its documented first step, and decide what to verify before adopting it.
Key takeaways
- Deploy on an isolated, pinned host with explicit network and credential boundaries.
- Keep client adapters optional and validate tool provenance separately.
- Use the documented Windows/Ubuntu regression and coherence gates for every routing change.
Choose an isolated host
The README lists Windows and Ubuntu CI, Kali and Ubuntu/Debian platform notes, Java/JDK, Node.js 22.12+, Python 3.x, and a compatible code AI client. Use a dedicated or disposable host for sensitive research, separate work cases from the repository, and keep real credentials out of environment examples.
Pin the reverse-skill tag or audited commit, inspect `README_AI.md`, `RULES.md`, and `docs/UV-AND-DOWNLOAD-SECURITY.md`, then record the host, client, tool versions, and permitted network profile.
Install and refresh deliberately
Clone the repository, run the platform refresh script, and inspect the generated tool index before loading a case. A tool appearing in the index does not prove its license, binary provenance, or suitability; attach checksums and ownership to the deployment receipt.
Keep optional client adapters outside the routing core. The README describes Claude Code, Codex, Cursor, and OpenCode as client-neutral integrations; test one adapter at a time so a client-specific permission does not silently widen the core scope.
Gate changes in CI
Routing/config changes should run the 173-case regression benchmark, structure and supply-chain coherence gate, smoke checks, and INDEX drift check on Windows and Ubuntu. Fail closed on unpinned auto-downloads or missing manifests.
For a fork, preserve the routing single source of truth, update generated indexes together, and publish the exact receipts and diffs. A green CI run proves consistency of the tested contract, not authorization for a new target.
Operate and recover
Back up routing manifests and redacted case metadata, rotate any service credentials, and define retention for samples, PCAPs, reports, and journals. If a tool or adapter is compromised, stop the case, quarantine artifacts, revoke credentials, restore a known-good commit, and rerun coherence checks.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Implementation steps
- 1
Choose a disposable host and pin reverse-skill plus runtime versions.
- 2
Refresh the tool index and review binary/license provenance.
- 3
Run routing, coherence, smoke, and index-drift checks in CI.
- 4
Back up manifests, define retention, and rehearse quarantine/rollback.
Copy-ready example
powershell -File skills/scripts/test-routing.ps1
powershell -File skills/scripts/verify-routing-coherence.ps1
powershell -File skills/scripts/smoke.ps1
powershell -File skills/scripts/extract-summaries.ps1 -CheckFrequently asked questions
Does CI validate every external tool's safety?
No. It validates the repository's routing and structure contract; inspect each tool, license, binary, and target authorization separately.
Can I put case files in the repository?
Keep local cases in the documented work area or a controlled evidence store, and commit only sanitized fixtures and manifests.
Sources
- reverse-skill README (captured 2026-08-31)Source checked 2026-08-31
- reverse-skill repositorySource checked 2026-08-31