log-analyzer
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE] (SAFE): The skill is entirely instructional and does not include any executable code, shell scripts, or binaries. It relies solely on the LLM's natural language capabilities.
- [DATA_EXFILTRATION] (SAFE): No network-capable commands (like curl or wget) or hardcoded credentials/secrets are present in the skill definition.
- [PROMPT_INJECTION] (SAFE): There are no markers or instructions designed to bypass agent constraints, reveal system prompts, or override core safety guidelines.
- [INDIRECT_PROMPT_INJECTION] (LOW):
- Ingestion points: The skill ingests untrusted log content provided by the user (SKILL.md).
- Boundary markers: Uses markdown code blocks for examples, but lacks explicit system-level instructions to ignore commands within the log data.
- Capability inventory: None. The skill has no access to subprocesses, file-writing, or network tools, limiting the potential impact of an injection to the content of the analysis itself.
- Sanitization: No input sanitization or validation logic is defined.
Audit Metadata