The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled Python script (scripts/redact_cast.py) to process .cast files via command-line invocation. This is the intended primary purpose of the skill and is implemented with multiple safety layers:
It requires separate input and output file paths to prevent accidental data loss.
It performs a verification scan on the generated output file to ensure all targeted patterns were successfully removed.
It uses regex-based redaction on a joined stream, effectively catching secrets that are typed one character at a time (keystrokes).
[DATA_EXPOSURE_&_EXFILTRATION]: The skill instructions explicitly warn against loading raw sensitive .cast files into the model context, which is a security best practice for handling potentially compromised recordings. It also recommends using a gitignored secrets.map file for literal replacements rather than passing secrets as command-line arguments, preventing them from appearing in shell history or process lists.
[PROMPT_INJECTION]: No malicious patterns, bypass attempts, or instructions to ignore safety guidelines were detected in the skill markdown or metadata.

cast-redact