The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by storing unvetted user input in documentation files that may be read by the agent in future sessions.
Ingestion points: User-provided content from the <content> argument is written to files in the .workflow/specs/ directory.
Boundary markers: Entries are structured using markdown headers and timestamps (e.g., ### [{type}] [{YYYY-MM-DD}]), providing structural separation but lacking explicit instructions to ignore potentially malicious content within the entry.
Capability inventory: The skill uses the Write tool to modify documentation files and the Bash tool to check for directory existence.
Sanitization: The skill validates the <type> argument against a whitelist of approved categories but does not sanitize the primary <content> payload.

spec-add