Skip to main content

Next steps

You already know the essentials: /rapid-init once per project, /rapid-<mode> for each job and rapid approve when the agent asks you to. From here, choose where to go next.

Practice with the labs​

Five small projects, one per mode, with the same four-step walkthrough. Start with Feature and continue with Bugfix. They are all in Labs.

Pick the right mode​

If you are going to…UseWhat Rapid OS requires
Build something new/rapid-featurePassing tests and your final approval
Fix a bug/rapid-bugfixA test that reproduces the bug before fixing it
Reorganize code without changing what it does/rapid-refactorThat the tests that were passing do not regress
Improve security/rapid-hardeningWork in an isolated copy and a human review
Investigate or compare options/rapid-researchFindings recorded; no tests required

Each mode has its guide in Engineering modes.

Work as a team​

  • Version the shortcuts with the repository so everyone uses the same workflow: rapid init --shortcuts at the project root leaves them in each agent's folder (.claude/skills/, .agents/skills/…).
  • After updating Rapid OS, repeat the command you used to install the shortcuts (rapid shortcuts install --global --harness <agent>, or rapid shortcuts install inside the project) to refresh the ones that changed.
  • In CI, the advanced mode commands verify the work without an agent: see CI/CD integration.

If something fails​

Rapid OS messages carry a RAPIDxxxx code. Look it up in the diagnostics catalog; the most common cases with shortcuts are in common problems.

Advanced mode​

You do not need it to work with shortcuts. If you want to understand what Rapid OS records at each step (contracts, evidence, evaluation), this is the full cycle with the v3 commands.

See the governance cycle under the hood

The full 8-phase cycle​

flowchart TD
    A[1. rapid scan] --> B[2. rapid spec]
    B --> C[3. rapid context]
    C --> D[4. rapid policy / run]
    D --> E[5. rapid harness resolve]
    E --> F[6. External Harness Execution]
    F --> G[7. rapid evidence add]
    G --> H[8. rapid eval run + validate]

The advanced governance stages​

1. Execution policies (rapid policy)​

Sets the team's corporate and risk rules in .rapid-os/policy.json: which risk levels (low, medium, high, critical) require isolated workspaces and which quality gates are indispensable.

2. Contracts and Run lifecycle (rapid run)​

Create a run with rapid run create --spec <spec-id> --harness <harness-id>. This generates an immutable ExecutionContract and opens the state ledger (RunState) to monitor task progress (T001...) and gate disposition (gate.tests...).

3. Capability resolution (rapid harness)​

Use rapid harness resolve to check whether the selected agent's profile declares support for the operations the contract requires. If the contract requires modifying files and the harness does not declare repository.write, the process flags the incompatibility before starting.

4. Physical evidence ingestion (rapid evidence)​

When the agent's work ends, collect the execution results (command logs, test reports, Git diffs) with rapid evidence add. The files are copied to .rapid-os/evidence/<run-id>/artifacts/ with their size and immutable SHA-256 hash.

5. Deterministic behavioral evaluations (rapid eval)​

Run rapid eval run --run <run-id>. The evaluation engine analyzes the collected physical evidence and issues a mathematical verdict (PASS, PASS_WITH_WAIVERS, UNVERIFIED or FAIL), guaranteeing that no task is considered complete without demonstrable proof.

To go deeper into each of these technical areas, see: