The Agent Skills Directory

[COMMAND_EXECUTION]: In search-cluster.py, the skill utilizes subprocess.run to execute a local helper script (stealth_fetch.py). The command execution is handled securely by passing arguments as a list rather than a shell string, and it uses a user-defined path for the Python interpreter.
[PROMPT_INJECTION]: The skill incorporates a native sanitization routine (internal_sanitize) that identifies and redacts keywords often used in prompt injection attempts within the retrieved search data. This acts as a safeguard against malicious content embedded in external web results.
[EXTERNAL_DOWNLOADS]: The skill communicates with several external search APIs and RSS feeds. It also requires the installation of the redis and scrapling Python libraries to function. All network operations use standard libraries and HTTPS.
[PROMPT_INJECTION]: The skill processes untrusted data from various search providers, creating an indirect prompt injection attack surface.
Ingestion points: Data retrieved from Wikipedia, Reddit, and DuckDuckGo (found in search-cluster.py).
Boundary markers: Absent; the data is returned to the agent context as JSON-formatted strings.
Capability inventory: The skill is capable of executing local subprocesses and performing outbound network requests.
Sanitization: Present; the internal_sanitize function redacts common injection patterns and strips non-printable characters from the input stream.

search-cluster