uipath-source-control-guardrails

Make authority boundaries executable so an agent can inspect and edit safely while high-consequence operations remain explicit human decisions.

Maturity: pilot.

Ownership Boundary

This custom skill owns: Make authority boundaries executable so an agent can inspect and edit safely while high-consequence operations remain explicit human decisions.

Keep current product commands, schemas, artifact validation, live tenant operations, and policy administration with official UiPath skills.

Workflow

1. Choose scope and authority

Determine project-only or user-level installation, the agent environment, commands the agent may run autonomously, and commands that require human execution or explicit per-action approval.

Completion criterion: A written command policy and installation target exist.

2. Install the hook

Choose Bash for Unix-like or WSL hooks and PowerShell for native Windows hooks. Copy the matching blocker into the supported hook location and merge its invocation into existing agent settings without replacing unrelated hooks.

Completion criterion: The guardrail is executable and connected to the intended command tool.

3. Customize patterns

Keep destructive Git patterns by default. Enable optional UiPath release-command blocking only when the team wants those commands guarded, and add organization-specific patterns sparingly.

Completion criterion: The block list matches the authority policy.

4. Verify blocked paths

Feed representative destructive commands to the hook on the target shell and observe a blocked exit and message for each policy class.

Completion criterion: Every guarded command class has a failing test.

5. Verify allowed paths

Test status, diff, log, read-only CLI, validation, and other approved commands so normal engineering work remains possible.

Completion criterion: Representative safe commands pass.

6. Document recovery

Explain how to run an approved operation manually, temporarily change policy through review, uninstall the hook, and recover from accidental partial operations.

Completion criterion: Users can distinguish protection from a broken tool.

Output Contract

• Command authority policy.
• Installed guardrail hook.
• Blocked and allowed command evidence.
• Recovery and maintenance instructions.

Guardrails

• Do not claim a hook protects environments it does not intercept.
• Do not hide the block list from repository maintainers.
• Do not block read-only or validation commands unnecessarily.
• Do not use the hook as a substitute for branch protection, least privilege, and platform authorization.

Example Requests

• "Prevent an agent from pushing or force-resetting this repository."
• "Require human execution for UiPath deploy and activate commands."
• "Add project-level source-control safety hooks."

Finish

Report completed work, observed evidence, the next official owner, and every blocker. Mark unobserved actions as pending.