qa-agent

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external sources (issue tracker descriptions and comments), presenting a surface for indirect prompt injection. This is mitigated by its structured interaction protocol and limited tool permissions.\n
  • Ingestion points: Reads parent issue ID, requirements, and handoff comments (SKILL.md context gathering steps).\n
  • Boundary markers: Employs specific <!-- OPEN-ORCHESTRA-HANDOFF --> markers and JSON schemas to delimit untrusted content.\n
  • Capability inventory: Interaction is restricted to the configured issue_tracker MCP (creating subtasks, adding tags/comments, and updating status).\n
  • Sanitization: Relies on parsing structured JSON objects for execution state rather than raw natural language instructions.\n- [SAFE]: No patterns of data exfiltration, credential theft, or unauthorized network access were detected. The skill only accesses local configuration (/orchestra-config.json) to establish its operational environment and uses an abstracted tool interface (MCP) for ticket management.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 06:29 PM