alarina
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill coordinates various specialized agents and processes potentially untrusted data from project repositories and user-provided diagnostic information, creating a surface for indirect prompt injection.\n
- Ingestion points: Data enters through the main request, existing project state/artifacts (SKILL.md), and runtime captures or traces (playbooks/investigation.md).\n
- Boundary markers: Present; references/coordination.md explicitly instructs the agent to treat worker outputs as evidence rather than instructions to prevent accidental command execution.\n
- Capability inventory: The skill manages agents like alaga and atona which have capabilities for file writing, code modification, and potential deployment coordination.\n
- Sanitization: The skill relies on instruction-based logic and role separation rather than technical input sanitization or escaping mechanisms.
Audit Metadata