The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data sources, creating a potential surface for indirect prompt injection attacks where malicious instructions could be hidden in processed content.
Ingestion points: The skill reads user-provided context, relevant project files, and data from external connectors (Google Search, web rendering services, and email service providers) as defined in SKILL.md and references/connectors.md.
Boundary markers: The instructions in SKILL.md and the Unified Skill Contract specifically warn the agent to treat evidence (scraped pages, reviews, emails) as untrusted input and to ignore any embedded instructions found within that data.
Capability inventory: Based on the provided files, the skill is limited to analysis and generating text-based artifacts; it does not exhibit high-privilege capabilities such as arbitrary shell execution or unauthorized file modifications.
Sanitization: The skill relies on instructional guardrails to prevent the AI from acting on untrusted input, which is a standard defensive measure for this category.

newsletter-monetization-planner