The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the exec tool to run a local bash script, scripts/byted-openclaw-diag.sh, which implements the diagnostic logic and log parsing.
[PRIVILEGE_ESCALATION]: In its "advanced mode," the skill has the ability to modify the openclaw.json system configuration file and restart the openclaw gateway service to enable debug logging. While these are administrative actions, they are explicitly associated with the tool's diagnostic purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from system log files and user session files, which could potentially contain instructions from past interactions.
Ingestion points: Logs and session files located in /tmp/openclaw/ and ~/.openclaw/agents/.
Boundary markers: The SKILL.md instructions include a "Direct Output Mode" that prevents the agent from summarizing the results, thereby reducing the risk of the log content influencing the agent's behavior.
Capability inventory: The skill has exec and read permissions.
Sanitization: The script uses Python for structured JSON parsing and includes a step to remove ANSI color codes using sed.
[DYNAMIC_EXECUTION]: The bash script uses python3 -c to execute inline code for processing JSON data and calculating metrics during execution.

byted-openclaw-diag