Contributing to Sentinel Scan CLI

Thanks for taking a look. This is a small, single-file tool, so contributions are easy to review and easy to make.

Reporting a false positive/negative

Use the Report a false positive/negative issue form. It asks for exactly what we need to reproduce and fix it: which check/OWASP tag fired, what endpoint type you scanned, what you expected vs what actually happened, and the sanitized scan output.

Proposing a new attack or heuristic

The attack corpus in sentinel_scan.py (ATTACKS) and the MCP heuristics are intentionally small and bounded so a scan stays fast and auditable. If you have a known prompt-injection technique or MCP manifest pattern that isn't covered, use the Missing attack / heuristic suggestion issue form, or:

1. Open an issue describing the technique and a source/reference if you have one.
2. If you'd like to submit a PR directly, add one (name, prompt) tuple to ATTACKS, keep the prompt text self-contained (no external calls), and explain in the PR description what class of attack it tests.

Development

No dependencies beyond the Python 3 standard library. Sanity-check any change with the built-in demo target before opening a PR:

python sentinel_scan.py --demo

Scope

This CLI is intentionally a fast heuristic smoke test, not a full audit engine (that's the paid Sentinel Scan audit). PRs that keep it dependency-free, fast, and easy to read in one sitting are the easiest to merge.