The Agent Skills Directory

[SAFE]: The skill is designed to improve agent security by enforcing restrictive CRM permission models. It includes local Python scripts for validation and auditing.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes potentially untrusted configuration files and directory contents.
Ingestion points: Files read by score_guardrails.py and directories scanned by skill_link_scan.py.
Boundary markers: Analysis is performed via specialized scripts rather than direct LLM interpolation.
Capability inventory: Read, Write, Glob, Grep, and Bash (strictly scoped to the skill's own scripts).
Sanitization: Inputs are sanitized through JSON parsing and regex matching.
[EXTERNAL_DOWNLOADS]: The skill provides links to documentation and GTM resources on jaymount.com. These references are auxiliary and are intended to be checked by the skill's own link scanning utility.

gtm-guardrails