The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted business data, client evidence, and account profiles, creating a potential surface for indirect prompt injection.
Ingestion points: Data enters the context via the setup profile, business economics, and evidence files loaded in step 1 of SKILL.md.
Boundary markers: The skill uses a structured 8-step execution path and refers to specific industry templates to maintain logical boundaries.
Capability inventory: The skill's primary capabilities include loading local reference files and invoking the related ads-meta skill for specific platform dimensions.
Sanitization: Extensive 'Required gates' are defined in the assets directory (e.g., assets/healthcare.md, assets/finance.md) which mandate privacy review, clinical-claim verification, and regulatory approval before processing high-stakes content.
[SAFE]: No malicious patterns such as remote code execution, credential exfiltration, or obfuscation were detected. The skill explicitly enforces data minimization and privacy isolation for client credentials and reports.

ads-plan