The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files to define its validation targets and parameters.
Ingestion points: The skill searches for and reads endpoints.json, tests/endpoints.json, and the repository's CLAUDE.md to discover API endpoints.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the values within the JSON configuration as untrusted data or to ignore embedded instructions.
Capability inventory: The skill utilizes the Bash tool for executing network requests, as well as Read, Write, and Edit tools for file manipulation.
Sanitization: While the skill warns against hardcoded IP addresses, it does not provide specific instructions for sanitizing or escaping the base_url, path, or headers values before they are interpolated into shell commands (e.g., curl).
[COMMAND_EXECUTION]: The skill is designed to generate and execute shell commands to perform sequential HTTP requests. This core functionality means that a malicious actor with the ability to modify configuration files in a repository could potentially influence the agent's network activity or attempt command injection through malformed URL paths or header values.

endpoint-validator