The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external web searches and project inbox entries to generate product specifications, which presents a surface for indirect prompt injection attacks.
Ingestion points: The process reads external Inbox Entries from the inbox/<repository>/ directory and performs 3–7 web searches to gather market data and competitive landscape information.
Boundary markers: The instructions do not include specific delimiters or guidance to ignore potentially malicious instructions embedded within the ingested inbox entries or search results.
Capability inventory: The skill has the ability to write files to the repository (docs/specs/), invoke other skills (business-analyst), and perform network operations via web tools.
Sanitization: There is no evidence of sanitization, validation, or filtering of the retrieved content before it is processed by the agent to draft the _idea.md specification file.

write-idea