mantis-report
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFE
Full Analysis
- Mandatory Secret and PII Redaction: The skill includes a dedicated instruction for the agent to scan and redact hardcoded API keys, tokens, credentials, and personally identifiable information (PII) before writing findings to the report. This is a proactive security control designed to prevent accidental data exposure in shared documentation.
- Snapshot Integrity and Provenance: The tool implements a robust tracking system for code snapshots, using commit hashes and status banners (e.g., mixed snapshots or dirty working trees). This ensures that findings are accurately linked to specific versions of the codebase, preventing confusion and maintaining the integrity of the security audit trail.
- Scoped File System Operations: The skill follows a 'state-relative' pathing convention, restricting all read and write operations to a dedicated workspace directory. It explicitly skips locator steps that would interact with the source code root during the reporting phase, adhering to the principle of least privilege.
- Structured Finding Deduplication: The reporting logic uses a strict 'same-bug predicate' based on UUIDs, signatures, and code paths to ensure findings are not lost or incorrectly merged. This logic is focused on data synthesis rather than execution, minimizing the attack surface during report generation.
Audit Metadata