The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from existing project files while possessing file system and shell execution capabilities. \n
Ingestion points: In SKILL.md (Phase 1), the agent reads top-level folder names and the contents of various project files, including manuscript drafts and notes. \n
Boundary markers: There are no explicit delimiters or instructions to ignore potential instructions embedded within the ingested project files. \n
Capability inventory: The skill frontmatter in SKILL.md enables the Write, Edit, and Bash tools, allowing the agent to modify the file system and execute shell commands. \n
Sanitization: No sanitization, filtering, or validation logic is specified for the external data before it is processed or used to generate actions.

intake-project