The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute uip CLI commands for project initialization (uip api-workflow init), validation (uip api-workflow validate), execution (uip api-workflow run), and deployment (uip solution publish). These commands are standard operations for managing the UiPath platform.
[DYNAMIC_EXECUTION]: The skill supports the creation and execution of JsInvoke activities, which run inline JavaScript within the workflow runtime. Additionally, it provides a Python script (score-evals.py) in the documentation to facilitate automated comparison of test results. These features are intended for developers building automation logic.
[INDIRECT_PROMPT_INJECTION]: The skill implements a 'loop mode' (Rule 22) where the agent autonomously runs test cases from a local evals/ folder and modifies the workflow until it passes. This creates a surface where malicious instructions embedded in test data could potentially influence the agent's code generation.
Ingestion points: The agent reads test case definitions (inputs and expected outputs) from evals/<scope>/eval-sets/*.json and processes raw execution logs from uip api-workflow run.
Boundary markers: The skill defines strict phases (Discovery, Plan, Generate, Validate) and requires explicit user consent for 'loop mode' and authenticated runs (Rules 21-22).
Capability inventory: The agent has access to Bash, Write, and Edit tools, enabling it to modify workflow JSON files and execute local build/run commands.
Sanitization: The agent is instructed to use structured JSON schemas (input.schema, output.schema) to define data contracts and uses standard JSON parsing in its scoring scripts.

uipath-api-workflow