The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/probe_debug.py) using the Bash tool to perform network scanning and analysis.\n- [DATA_EXFILTRATION]: The skill performs network requests to target URLs provided by the user to check for exposed debug endpoints. It includes a mandatory --authorized flag for non-local targets to ensure the user attests to having permission for the scan.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes response bodies from external web servers during its scanning process.\n
Ingestion points: The response body is read in scripts/probe_debug.py using the safe_get utility.\n
Boundary markers: The skill does not use specific delimiters for external content in its output, but it performs analysis using regex matching on a limited portion of the response body (first 4KB) rather than direct prompt interpolation.\n
Capability inventory: The skill uses safe_get for network requests and outputs findings through the lib.report module.\n
Sanitization: The script performs regex fingerprinting to validate that the response matches expected patterns for administrative panels rather than arbitrary content.

detecting-debug-endpoints