The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted external data from the project repository to infer business purposes, compliance requirements, and risk thresholds. Ingestion points: The workflow involves scanning project metadata, configuration files, database migrations, API specifications, and documentation (e.g., README, package.json, pom.xml, OpenAPI specs) in the SKILL.md file. Boundary markers: There are no instructions provided to the agent to treat content within these files as untrusted or to ignore embedded instructions. Capability inventory: The agent performs extensive file reading and logical tracing (e.g., tracing PII handling and payment flows) across the codebase. Sanitization: The skill lacks requirements for sanitizing or validating the content of the files before the agent processes them.

pasta-objectives