session-logs-hardened

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill facilitates access to local session logs in ~/.openclaw/agents/, which contain sensitive historical conversation transcripts. To mitigate exposure and exfiltration risks, the instructions explicitly forbid piping this data to network-active tools like curl or ssh.
  • [PROMPT_INJECTION]: The skill processes untrusted historical conversation data, which presents a surface for indirect prompt injection.
  • Ingestion points: Reads message content from JSONL session files located in ~/.openclaw/agents/<agentId>/sessions/ (SKILL.md).
  • Boundary markers: No explicit delimiters are used in the provided command templates, although the skill includes instructions for the agent to verify search scope with the user.
  • Capability inventory: Uses local bash tools including jq, rg, ls, grep, and awk for data processing (SKILL.md).
  • Sanitization: Content is processed without specific programmatic sanitization or filtering of embedded instructions.
  • [COMMAND_EXECUTION]: The skill uses several common shell utilities (jq, rg, awk) to extract and summarize session data. These commands are executed locally and are integral to the skill's primary function of log analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 06:32 PM
Security Audit — agent-trust-hub — session-logs-hardened