The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources including web scrapes, reviews, and comments, which presents a surface for indirect prompt injection.
Ingestion points: The skill explicitly mentions reading scraped page text, reviews, comments, and emails as data sources in SKILL.md.
Boundary markers: The instructions in SKILL.md and references/skill-contract.md contain strong directives to treat external material as untrusted evidence and to not follow any instructions embedded within that data.
Capability inventory: While primarily analytical, the skill references potential integration with external marketing and search tools in references/connectors.md.
Sanitization: The primary defense mechanism is the explicit natural language instruction to the agent to ignore instructions found within external evidence.

competitor-tracker