evidence-selfloop
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs standard file processing tasks. It reads intermediate research artifacts and writes a diagnostic report to the output directory.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external JSONL files (
outline/*.jsonl,papers/*.jsonl). While this data could theoretically contain malicious instructions, the skill's logic is limited to aggregation and report generation, and the risk is inherent to the data-processing nature of the tool. - Ingestion points: Multiple JSONL files in
outline/andpapers/directories within the provided workspace. - Boundary markers: Not present; the data is treated as trusted pipeline output.
- Capability inventory: File reading and writing via local filesystem.
- Sanitization: Relies on
json.loadsfor structural parsing; text content is processed as-is.
Audit Metadata