skills/insight68/skills/session-logs/Gen Agent Trust Hub

session-logs

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill accesses sensitive conversation transcripts stored in the ~/.openclaw/agents/<agentId>/sessions/ directory. These files contain full historical records of past interactions.
  • [COMMAND_EXECUTION]: The skill provides numerous templates for executing shell commands such as jq, rg, awk, and sort to process local session log files.
  • [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it reads and re-introduces historical conversation data into the agent's context. If a previous session contained malicious instructions, the agent could be influenced when analyzing those logs.
  • Ingestion points: Local conversation data is ingested from JSONL files in ~/.openclaw/agents/<agentId>/sessions/.
  • Boundary markers: No specific boundary markers or safety instructions are defined in the command templates to isolate log content from active instructions.
  • Capability inventory: The skill utilizes local file reading and shell-based data processing tools.
  • Sanitization: No sanitization or filtering of the extracted text content is described or implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 08:21 PM