The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in tests/test_scripts.py to execute internal validation scripts (audit_edit_boundaries.py and validate_delivery.py) during testing. This is a standard practice for automated testing of local tools and does not involve external or untrusted input.
[REMOTE_CODE_EXECUTION]: SKILL.md contains bash instructions for running local telemetry scripts if they exist (python3 telemetry/version_check.py). While this is a form of local script execution, it is scoped to the repository's own files and includes explicit privacy-preserving and opt-in guidance for the user.
[DATA_EXFILTRATION]: The skill mentions optional telemetry for version checks, but explicitly instructs the agent to 'Never log content, URLs, paths, credentials, names, or business data.' No hardcoded exfiltration patterns or network operations to non-whitelisted domains were found.
[PROMPT_INJECTION]: The skill instructions emphasize 'source-grounded' analysis and include strict rules for obtaining user approval before any publication or sharing actions. There are no attempts to override safety filters or bypass constraints.

leveling-up-content-engine