The Agent Skills Directory

[SAFE]: The skill is a legitimate text generation tool designed to ground its output in project-specific data. It reads local context files such as .ulpi/launch/positioning.md and visits product websites using the browse tool, which is expected behavior for its primary purpose.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted data from external websites. Evidence: (1) Ingestion points: Live websites via the browse tool and local files like project-context.md. (2) Boundary markers: Absent. (3) Capability inventory: Read, Write, AskUserQuestion, and Skill invocation. (4) Sanitization: Absent, though the skill contains strict logic-level grounding rules and a self-check step to ensure accuracy. This is a low-risk surface given the skill's specific use case.

launch-copy