session-logs
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides shell command templates, including loops and pipes, for the agent to execute in order to parse and summarize JSONL log files using tools like jq, rg, and awk.
- [PROMPT_INJECTION]: The skill processes historical conversation logs which could contain malicious instructions from previous sessions, creating an indirect prompt injection surface.
- Ingestion points: Local session logs located at ~/.clawdbot/agents//sessions/*.jsonl.
- Boundary markers: Absent; commands extract and return raw text content from the logs to the agent.
- Capability inventory: Execution of shell-based parsing and analysis tools.
- Sanitization: No sanitization or filtering of the content within the log files is performed before the agent processes it.
Audit Metadata