The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The script scripts/lighthouse_audit.py uses subprocess.run with a list of arguments to invoke the lighthouse binary. This pattern is resistant to shell injection because it avoids shell interpretation of the input parameters.
EXTERNAL_DOWNLOADS (LOW): The skill depends on the lighthouse CLI. As a widely recognized tool from a trusted organization (Google), the dependency risk is downgraded to LOW per the [TRUST-SCOPE-RULE].
Indirect Prompt Injection (LOW): The skill ingests data from external URLs, creating a potential surface for indirect prompt injection.
Ingestion points: scripts/lighthouse_audit.py fetches and parses content from user-provided URLs.
Boundary markers: Absent. The resulting audit data is passed to the agent without delimiters or specific instructions to treat the content as untrusted.
Capability inventory: Access to Bash and the ability to execute system commands via subprocess in scripts/lighthouse_audit.py (line 25).
Sanitization: Absent. The script extracts scores and summaries from the external report without filtering for instruction-like patterns.

performance-profiling