The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and validation reports include references to executing local Python scripts for linting and cluster validation. These are intended as internal quality control mechanisms. 1. Evidence in SKILL.md: python3 ./plans/seo-llm-skill-cluster/scripts/lint_skill_cluster.py .. 2. Evidence in validation-report.md: Mentions of lint_production_skill.py, quick_validate.py, and lint_skill_cluster.py used to verify skill state.
[PROMPT_INJECTION]: The skill ingests untrusted data from user requests and external website audits, which constitutes an attack surface for indirect prompt injection. This is addressed through a structured workflow and a specific refusal policy for unethical behavior. 1. Ingestion points: User requests for site planning, live HTML audits, schema verification, and crawler log monitoring (mentions in SKILL.md and references/workflow.md). 2. Boundary markers: The skill utilizes a 'Handoff Packet' template (assets/handoff-packet.template.md) to structure internal communication, but lacks explicit delimiters for raw external website content. 3. Capability inventory: The skill can trigger local Python script execution and browser-based research via supporting skills. 4. Sanitization: Relies on explicit 'Safety and Refusal Rules' in SKILL.md to guide the agent in identifying and rejecting malicious or deceptive SEO requests.

site-growth-orchestrator