The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes information from external documentation websites to generate configuration rules. Ingestion points: Documentation URLs for various agents provided in references/puntos-de-partida.md and user-supplied rules in SKILL.md. Boundary markers: No explicit delimiters are specified for handling the ingested documentation content. Capability inventory: The skill performs file writes to agent configuration files and executes terminal commands. Sanitization: The workflow requires a mandatory human review and approval step (Step 4) before any configuration is applied, mitigating the risk of malicious instructions from external sources.
[COMMAND_EXECUTION]: The skill executes terminal commands to verify the newly established guardrails. Evidence: Step 6 of the workflow in SKILL.md involves triggering the regulated event (e.g., git push) to confirm the guardrail is operational.

crear-guardrail