ops-inspector
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains proactive defensive instructions to prevent exploitation via log content. It explicitly directs the agent to treat log content as passive data and ignore any embedded command patterns such as 'ignore previous instructions'.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a diagnostic surface for indirect prompt injection through log analysis.
- Ingestion points: Environment logs and function outputs are fetched via
queryLogsandqueryFunctions(action="listFunctionLogs")in SKILL.md. - Boundary markers: The skill provides explicit natural language boundary markers, instructing the agent that log content must not be interpreted as instructions.
- Capability inventory: The skill relies on platform-native, read-only diagnostic tools.
- Sanitization: Security is enforced via model-level instructions to maintain strict separation between data and instructions.
Audit Metadata