bug-triage
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted bug report files which could contain malicious instructions.
- Ingestion points: Reads files from
production/qa/bugs/*.md,production/qa/bugs.md, andproduction/qa/qa-plan-*.md. - Boundary markers: The skill formats its own output as a structured triage report with specific headers, helping to distinguish processed data from original inputs.
- Capability inventory: The skill has
WriteandEditpermissions but only writes to a specific report file path (production/qa/bug-triage-[date].md) after explicit user approval. - Sanitization: The skill uses classification logic (Severity S1-S4, Priority P1-P4) to transform input data into specific categories, which naturally filters out non-compliant instructions in the final report. The Collaborative Protocol explicitly forbids auto-closing bugs or auto-assigning without user review.
Audit Metadata