> ## Documentation Index
> Fetch the complete documentation index at: https://plainrouter.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Review action proposals

> Evaluate evidence, policy checks, execution consequences, and verification state.

Open **Approval queue** from the PlainRouter sidebar to review agent proposals for the current workspace.

## Read the queue summary

The page shows the current policy mode and four counts:

| Count                     | Meaning                                                                                                                            |
| ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| **Waiting**               | Batches awaiting a human decision.                                                                                                 |
| **Awaiting verification** | Provider writes that are still **Not Landed** while durable receipt-addressed verification continues.                              |
| **Landed**                | Batches whose executed actions all have verified provider receipts. Suggest-only approvals and measurement-only runs are excluded. |
| **Blocked**               | Batches stopped by proposal-time or execution-time policy.                                                                         |

Use **Pending** for decisions that need you. Use **History** for approved, rejected, blocked, executing, awaiting-verification, failed, completed, and rolled-back outcomes.

## Review a proposal card

### Proposer and account

The header identifies the agent, Meta ad account, proposal time, and current batch status. The card and approval control reflect the frozen execution-mode snapshot. Confirm the account is the one you intended to authorize.

### Why now

The rationale should describe the evidence, proposed response, and expected outcome. It must not present a human-supplied target or aggregate Signals data as current provider state.

### Evidence provenance

Each cited source shows its scope, exact fields, observation time, and measurement window where applicable.

For creative actions:

* `get-creative-library` provenance proves selected-asset identity, approved-account ownership, readiness, and status.
* `staged-asset-manifest` provenance proves the upload uses frozen bytes with a specific SHA-256 digest and detected content type.
* Creative performance is excluded from the durable action-bound citation.

### Dynamic policy checks

Provider-backed fields may be **Available**, **Partially available**, or **Unavailable**. Each missing field is marked as blocking or disclosed but non-blocking.

Executable work fails closed when a provider-required or active-policy-required field is unknown. In **Suggest only**, unavailable state may still be presented for human review because approval cannot execute it.

### Action and consequence

Confirm every target and parameter. The approval control states the consequence:

| Button                          | Result                                                                                              |
| ------------------------------- | --------------------------------------------------------------------------------------------------- |
| **Approve recommendation**      | Records **Approved without execution**. No Meta write occurs.                                       |
| **Approve and execute**         | Authorizes the executable snapshot and queues a fresh policy check before provider execution.       |
| **Approve one-run measurement** | Starts a closed diagnostic run. It is explicitly **Not Landed** and requires verified compensation. |

On mobile, the swipe control also states whether approval executes.

## Approve executable creative work

<Steps>
  <Step title="Confirm the approved account">
    Verify the account name and platform-native target IDs.
  </Step>

  <Step title="Confirm evidence and policy">
    Read each citation, any unavailable fields, protected-entity checks, and the proposal-time mode snapshot.
  </Step>

  <Step title="Confirm permanent pause">
    For an ad copy, verify the proposal uses `status: paused`. This release cannot activate the copy.
  </Step>

  <Step title="Approve">
    Click **Approve and execute**, or use the mobile approval control. Approval is authoritative once and does not bypass the execution-time policy check.
  </Step>

  <Step title="Follow verification">
    Open **History**. A successful immediate verification becomes **Completed** and increases **Landed**. An `IN_PROCESS` copy becomes **Executed pending verification** and shows its receipt, latest state, next scheduled read, cutoff, and safety deadline.
  </Step>
</Steps>

<Warning>
  Do not interpret **Approved**, **Executing**, or **Executed pending verification** as Landed. Landed requires exact provider verification for every action in the batch.
</Warning>

## Reject a proposal

Click **Reject**, optionally explain what the agent should reconsider, and confirm **Reject proposal**. The reason is stored in the audit event.

Approval and rejection are idempotent terminal decisions for a pending batch. A second decision does not reopen it.

## Review checklist

* Is this the correct Meta ad account?
* Did a person intend each source ad and asset selection?
* Are evidence claims limited to the recorded fields and scope?
* Does the selected mode make this approval non-executing or executable?
* Is every missing critical field handled as a block?
* Is a copied ad permanently `paused`?
* If work is awaiting verification, do you understand it is Not Landed?
* Does any rollback-incomplete state require operator attention?