Skills
skills/bytedance/agentkit-samples/byted-openclaw-diag/Gen Agent Trust Hub

byted-openclaw-diag

Warn

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/byted-openclaw-diag.sh executes administrative commands such as openclaw gateway restart to apply diagnostic settings and uses python3 -c one-liners to dynamically modify values in the openclaw.json configuration file.
  • [CREDENTIALS_UNSAFE]: The skill accesses highly sensitive session data stored in ~/.openclaw/agents/*/sessions/*.jsonl, which contains the complete history of user prompts, agent reasoning, and tool outputs. This constitutes a significant data exposure risk as it reveals private user interactions.
  • [COMMAND_EXECUTION]: The diagnostic script performs file system operations including cp to create backups of configuration files and rm for cleaning up diagnostic artifacts.
  • [DATA_EXFILTRATION]: The skill defines a 'direct output' mode that bypasses the model's summarization and safety filtering by instructing the agent to send the raw diagnostic output (potentially containing sensitive logs or PII) directly to the user using the message tool.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 31, 2026, 02:18 PM
Security Audit — agent-trust-hub — byted-openclaw-diag