Allowed action
Supply an action whose tool, target, purpose, identity context, and policy match an explicit rule. Confirm the policy disposition is PERMIT before the host executes.
5D setup skill
Set up, test, and maintain one action boundary.
Use the 5D Skill to inspect your deployment, configure a documented action path, test GREEN/YELLOW execution eligibility, ORANGE hold, and RED block, then repeat verification whenever policy, tools, or integration paths change.
Reproducible path
Install the latest stable package.
Tested release: 0.7.0. The compatibility evidence is tied to that release; the install command retrieves the latest stable version.
python -m pip install fivedriskBlocked action
Supply an action matching a declared prohibition. Confirm the disposition is BLOCK and that the host does not call the tool.
Unresolved outcome
Simulate a lost downstream response or missing evidence. Keep the state unknown; do not treat a timeout as success, denial, or permission to retry blindly.
Before production
Test the surrounding application too.
- Put the gate on every relevant execution path.
- Validate identity, policy version, destination, and action context in the host application.
- Define idempotency, reconciliation, reviewer availability, and recovery for each tool.
- Use the worked examples and evidence records to separate demonstrated behavior from assumptions.
5D setup skill
A reviewable implementation accelerator.
The complete skill is hosted in the public GitHub repository. It is not bundled in the 0.7.0 PyPI wheel. Give the repository-hosted skill to your coding agent to inspect the project, select a documented integration path, scaffold policy, and run positive and negative checks. Read every proposed change before accepting it. A successful skill run does not certify the application or prove that alternate execution paths are covered.