The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a clear ingestion surface (web page content, scripts, and metadata). However, it includes robust defensive instructions, mandating that the agent treat all external data as untrusted and specifically instructing it not to collect or follow hidden instructions found within that data.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions proactively mitigate data exposure by explicitly forbidding the collection of credentials, form data, analytics identifiers, or customer data. It uses local project directories for storage, following least-privilege principles.
[SAFE]: No malicious patterns, obfuscation, or unauthorized remote execution were detected. The skill's stated purpose aligns with its functional instructions.

ads-dna