The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and displays data from external API responses, which are untrusted sources that could contain malicious instructions designed to influence the agent.
Ingestion points: API response bodies are captured and printed to the console in SKILL.md (Step 2) and scripts/agent.py (via the test_unauthenticated_endpoints method).
Boundary markers: The skill does not employ delimiters or specific instructions to the agent to treat external response data as untrusted text.
Capability inventory: The skill possesses network communication capabilities (requests) and file system write access (agent.py generates a JSON report).
Sanitization: No sanitization, filtering, or escaping is performed on the content retrieved from external API endpoints before it is processed or displayed.
[COMMAND_EXECUTION]: The SKILL.md and references/api-reference.md files provide instructions for using external command-line tools like hashcat, Hydra, and jwt_tool. Although these are not directly invoked by the script, the instructions prompt the agent/user to perform external command execution for exploitation purposes.
[DATA_EXFILTRATION]: The skill's endpoint scanner in both SKILL.md and scripts/agent.py explicitly probes for sensitive file paths and configurations, including /.env, /config, and /actuator/env, which are common targets for credential harvesting and information disclosure.

testing-api-authentication-weaknesses