The Agent Skills Directory

[SAFE]: No malicious patterns, unauthorized network operations, or credential exposures were detected in the skill instructions or reference files. The skill includes explicit guidance for secure credential management by advising that API keys be kept out of source control.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the inherent risk of processing data from external, untrusted sources and establishes robust defensive protocols.
Ingestion points: The skill is designed to process external context including scraped web text, reviews, comments, and emails, as specified in SKILL.md and references/skill-contract.md.
Boundary markers: All findings and claims must be explicitly labeled with provenance and confidence markers (e.g., measured, user-provided, estimated) to distinguish them from direct instructions.
Capability inventory: The skill's scope is primarily analytical; any persistent writes, account mutations, or external side effects are strictly governed by the Unified Skill Contract and require explicit human authorization before execution.
Sanitization: The instructions contain strong defensive prompts, explicitly directing the agent to treat external content solely as evidence and to disregard any instructions or commands embedded within that data.

marketing-psychology