The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data from potentially untrusted sources such as scraped web pages, customer reviews, and emails. This creates a surface for indirect prompt injection attacks; however, the skill includes explicit defensive instructions to treat these inputs as untrusted and to disregard any embedded instructions within the data.
Ingestion points: Identified in the 'Data Sources' section of SKILL.md and the 'Core Invariants' of references/skill-contract.md.
Boundary markers: The skill mandates that embedded instructions from evidence be ignored.
Capability inventory: The skill uses tools for search and marketing automation, but requires user authorization for all side effects.
Sanitization: The agent is instructed to label all evidence by provenance and treat third-party content as data, not instructions.
[SAFE]: The skill implements a comprehensive safety model through its 'Unified Skill Contract' and 'Routing Policy'. These documents establish clear rules for evidence precedence, data handling, and permission management, ensuring the agent operates within a bounded and transparent scope.

launch-asset-packager