The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external API responses, which could potentially contain malicious instructions intended to influence the agent.
Ingestion points: Data is ingested via requests.get(...).json() in scripts/agent.py and through curl output in SKILL.md.
Boundary markers: There are no delimiters or explicit instructions provided to the agent to treat external API content as untrusted.
Capability inventory: The agent has the capability to perform network requests and write reports to the local file system.
Sanitization: No sanitization or validation of the retrieved API content is performed before processing or display.
[REMOTE_CODE_EXECUTION]: The skill documentation includes examples of shell commands that pipe remote content directly into a Python interpreter.
Evidence: SKILL.md contains the command curl -s https://target-api.example.com/api/docs/swagger.json | python3 -m json.tool.
Analysis: Although this instance is used for JSON formatting, piping remote data into a language interpreter is a risky practice that can lead to arbitrary code execution if the source URL is compromised or the command is modified.
[COMMAND_EXECUTION]: The skill relies on shell commands for API discovery and testing, executing external utilities like curl and python3 using user-provided parameters.

testing-api-for-broken-object-level-authorization