The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of API access logs, creating a potential surface for indirect prompt injection.
Ingestion points: The parse_access_log function in scripts/agent.py and the EnumerationDetector class in SKILL.md process data directly from external log files.
Boundary markers: The skill uses regular expressions to structure the data but lacks explicit delimiters or instructions for the agent to ignore potential commands embedded within the log content.
Capability inventory: The skill performs local file reading and utilizes the requests library to make network queries to a WAF API based on user-provided parameters.
Sanitization: While the script uses regex to extract specific fields, the resulting strings are not sanitized to prevent them from being interpreted as instructions if the analysis output is further processed by an LLM.

detecting-api-enumeration-attacks