The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses a sensitive credential file located at '$HOME/.clawdbot/credentials/unraid/config.json' to retrieve API keys for Unraid server authentication. Additionally, the 'scripts/unraid-query.sh' helper script uses the 'curl -k' (insecure) flag, which disables SSL certificate verification, potentially exposing authentication tokens and sensitive data to man-in-the-middle attacks.
[DATA_EXFILTRATION]: The 'scripts/dashboard.sh' script aggregates sensitive system information—including hardware specifications, storage capacity, cache pool status, and recent system logs from multiple servers—into a single unencrypted markdown file at '$HOME/clawd/memory/bank/unraid-inventory.md'. This concentration of sensitive information in a predictable local path increases the risk of data exposure.
[COMMAND_EXECUTION]: The 'scripts/unraid-query.sh' script is susceptible to JSON parameter injection. It constructs the API request by interpolating the '$QUERY' variable into a double-quoted JSON string in a 'curl' command without proper escaping. This allows a crafted input to manipulate the structure of the GraphQL request. Furthermore, the 'examples/read-logs.sh' script dynamically generates these queries using un-sanitized arguments.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it retrieves and processes system log content from external Unraid servers.
Ingestion point: The 'examples/read-logs.sh' script fetches data from the 'logFile' GraphQL endpoint.
Boundary markers: None; log content is returned directly to the agent's context.
Capability inventory: The skill can execute shell commands via 'curl' and 'jq', perform network requests, and read local configuration files.
Sanitization: No sanitization or escaping is performed on the log data before it is presented to the agent.

unraid