The Agent Skills Directory

[PROMPT_INJECTION]: The skill's workflow establishes an indirect prompt injection surface by requiring the agent to process potentially untrusted data to generate its plans.
Ingestion points: The agent is instructed to read local files, including code, configurations, logs, and issue text, as well as external 'official documentation' and 'vendor docs' (SKILL.md).
Boundary markers: The skill lacks technical delimiters to isolate ingested data from agent instructions, although it provides procedural rules to treat user claims as 'unproven' and maintain original intent (SKILL.md).
Capability inventory: The agent is granted the ability to write Markdown artifacts to the local workspace (SKILL.md).
Sanitization: There are no instructions for sanitizing or escaping content from external or local sources before it is interpolated into the generated plans.

vibe-planning