soia-dev-fix-loop
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill references an external resource from the vendor's own namespace (
soia-team/soia-open-dev-coding-skills) for installation, which is a legitimate vendor resource. - [COMMAND_EXECUTION]: The fix loop workflow involves executing shell commands to reproduce reported bugs and run regression tests. These actions are bounded by the skill's primary purpose of software development and maintenance within the target workspace.
- [PROMPT_INJECTION]: The skill processes untrusted external data in the form of 'findings' (review comments or test logs) to drive code modifications and verification. This constitutes an indirect prompt injection surface.
- Ingestion points: External findings and review comments ingested in step 1 of the loop.
- Boundary markers: The skill utilizes structured markdown tables to normalize findings, though it lacks explicit instructions to ignore potentially malicious embedded commands.
- Capability inventory: The skill performs file writes (implementing fixes) and subprocess execution (tests and reproduction) across the workflow.
- Sanitization: Mitigation relies on a process-oriented approach, requiring evidence-based fixes, independent review, and regression testing to verify that changes match intended behavior.
Audit Metadata