The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of untrusted external data (code repositories), which can contain instructions disguised as code or comments. Ingestion points: The skill takes 'Code repo(s) to analyze' as input for mapping existing systems (found in skill.md). Boundary markers: There are no instructions provided to the agent to treat external code as data only or to ignore potential embedded instructions, which could lead to obedience to malicious prompts hidden in code. Capability inventory: The skill requires the agent to read, parse, and reason about file contents across an entire codebase, providing a broad attack surface. Sanitization: No sanitization or filtering of external input is mentioned in the procedures.

breadboarding