The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill retrieves historical API specifications and metadata from the Wayback Machine (web.archive.org), which is a well-known service for digital archiving and research.
[COMMAND_EXECUTION]: The skill utilizes standard system utilities to automate security testing workflows. This includes curl for network requests and API probing, jq for parsing and extracting data from JSON-formatted OpenAPI/Swagger specifications, and comm for identifying differences between endpoint inventories.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection due to its processing of external data.
Ingestion points: The skill fetches OpenAPI/Swagger specifications from external target servers and the Wayback Machine (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore instructions that might be embedded within the retrieved JSON specifications.
Capability inventory: The skill uses curl for network operations, jq for data processing, and bash loops for automated interaction with discovered endpoints.
Sanitization: Endpoint paths extracted from external JSON files are interpolated into shell commands without explicit validation or sanitization, creating a potential surface for influencing agent behavior if a malicious specification is processed.

hunt-shadow-api