The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided design documents (sdd.md), which represents an indirect prompt injection surface.
Ingestion points: references/planning.md (Step 2) reads the user's Markdown specification.
Boundary markers: The skill enforces multiple hard stops requiring explicit user approval via AskUserQuestion. Specifically, the generated tasks.md plan must be approved before implementation (Rule 9), and a second mandatory gate exists between structural building and detail wiring (Rule 26).
Capability inventory: The skill uses Bash to execute the uip CLI and Write/Edit to modify local Case Management JSON files.
Sanitization: The skill relies on human review of the generated declarative plan (tasks.md) and the project skeleton in Studio Web to detect and prevent malicious behaviors.
[EXTERNAL_DOWNLOADS]: The skill contains logic to install or upgrade the @uipath/cli package from the public npm registry.
Evidence: references/planning.md (Step 0) checks the local version and executes npm install -g @uipath/cli@latest if necessary.
Trust Context: The download originates from a well-known public registry and is the official tool provided by the skill's author, aligning with the primary purpose of authoring Case Management projects.

uipath-case-management