orch-fix-defect
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted bug reports to drive code reproduction and modification tasks, which could be exploited to influence agent behavior. * Ingestion points: External defect descriptions and reproduction steps provided to the agent (SKILL.md). * Boundary markers: No delimiters or ignore instructions are specified in the instructions to protect against instructions embedded in defect reports. * Capability inventory: Bug reproduction (command execution via tests), file modification (writing code fixes), and automated git operations (committing code at Gate 2). * Sanitization: No sanitization, validation, or filtering of the input content is described in the workflow.
Audit Metadata