> For the complete documentation index, see [llms.txt](https://cerebral-systems.gitbook.io/cerebral-systems-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cerebral-systems.gitbook.io/cerebral-systems-docs/architecture/verdicts-and-approvals.md).

# Verdicts and approvals

Every proposal submitted to `POST /v1/actions` returns a **verdict**. The verdict is the whole decision: what Warrant decided, why, and what happens next.

**The three decisions**

| Decision         | Meaning                                                                     | What happens                                                                                                           |
| ---------------- | --------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------- |
| `allow`          | The action is within policy and below the approval threshold.               | Warrant enqueues a durable execution job and runs it through the registered actuator. Final status becomes `executed`. |
| `needs_approval` | The action is permitted but its risk is at or above `approval_required_at`. | The action is **held** (`pending_approval`) and routed to `review_route`. Nothing executes until a human approves.     |
| `deny`           | The action violates policy.                                                 | Nothing runs. `blocking_reasons` explains which check failed.                                                          |

A verdict also carries the individual `checks` (allowlist, operation allowlist, parameter constraints, risk), the `effective_risk` Warrant derived, the `authenticated_principal`, the policy and catalog commitments (hashes), and an `audit_ref`. The `actor` and `risk` you submitted are recorded as claims — Warrant derives authoritative identity from the API key and may raise risk from a reviewed catalog.

**Action lifecycle**

```
proposed ─▶ verdict
              ├─ allow          ─▶ execution_pending ─▶ executed | execution_failed
              ├─ needs_approval ─▶ pending_approval ─▶ (approve) ─▶ execution_pending ─▶ executed
              │                                     └─ (reject)  ─▶ rejected
              └─ deny           ─▶ denied
```

**Approving a held action**

Held actions appear in the application's **Approvals** tab in the dashboard, where a reviewer can approve or reject with a note. To automate it, call the approval API with an **approver-scoped** key (never the agent key):

```sh
curl -sS "$WARRANT_URL/v1/actions/<action-id>/approve" \
  -X POST \
  -H "Authorization: Bearer $WARRANT_APPROVER_KEY" \
  -H "Content-Type: application/json" \
  -d '{"note": "reviewed — staging only"}'
```

Use `/reject` to decline. Approving records the authenticated approver, then enqueues the same execution job an `allow` would have. An agent key cannot approve — the approval API rejects it — which is what keeps the human in the loop.

**How agents should react**

| Result                     | Agent behavior                                                      |
| -------------------------- | ------------------------------------------------------------------- |
| `allow`                    | Done — Warrant owns execution. Do not call the provider yourself.   |
| `needs_approval`           | Stop the tool turn; surface the action id and `review_route`.       |
| `deny`                     | Stop; surface the `blocking_reasons`.                               |
| HTTP `202` + `polling_url` | Execution is still running; poll — never duplicate the side effect. |

**Audit trail**

Every action accumulates a tamper-evident trace: `proposal_received`, `verdict`, `approval` (if held), and `execution`. Fetch it with `GET /v1/actions/<action-id>`, or `GET /v1/actions/<action-id>/audit-export` for the Merkle-rooted evidence bundle, whose `verified` flag confirms the chain has not been altered.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://cerebral-systems.gitbook.io/cerebral-systems-docs/architecture/verdicts-and-approvals.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.