The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill explicitly manages and provides instructions for retrieving sensitive data such as gateway.auth.token and models.<provider>.apiKey via the openclaw config get command as documented in references/config-paths.md.
[COMMAND_EXECUTION] (HIGH): The skill is granted Bash(openclaw *) permissions, allowing it to execute powerful administrative subcommands including doctor --fix and security audit --fix which modify system state and file permissions (chmod).
[DATA_EXFILTRATION] (HIGH): The ability to retrieve configuration values poses a critical risk. If an attacker can influence the agent's input, they could trick it into executing openclaw config get gateway.auth.token and revealing the secret in the output.
[PROMPT_INJECTION] (HIGH): This finding identifies a high vulnerability surface for Indirect Prompt Injection.
Ingestion points: The skill processes untrusted user input via the <what to fix or configure> argument to determine which administrative actions to take.
Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands in the user input.
Capability inventory: The skill has the capability to modify global system configurations, change file permissions, and restart services.
Sanitization: Absent. The skill relies on the agent's interpretation of the user's natural language request to select and execute bash commands.

openclaw-admin