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

# Remediation Actions

> Understand available security actions and the approval workflow

Hiro can propose and execute remediation actions across your connected integrations. All high-impact actions require explicit approval, ensuring you maintain full control over your security response.

## How Remediation Works

When Hiro identifies a threat, it:

1. **Analyzes the situation** — Correlates evidence from multiple sources
2. **Determines appropriate actions** — Based on threat type and severity
3. **Calculates confidence** — How certain it is the action is warranted
4. **Proposes the action** — Presents it for your review
5. **Awaits approval** — Executes only after you confirm
6. **Logs the result** — Maintains a complete audit trail

<Warning>
  Hiro will never execute a destructive action without explicit approval. You always see what will happen before it happens.
</Warning>

## Approval Levels

Hiro uses a tiered approval system:

### No Approval Required

Read-only operations that don't modify state:

* Querying logs and events
* Listing users, groups, and resources
* Reading configuration
* Fetching audit trails

### Approval Required

Actions that modify state or could impact users require explicit approval before execution. Both Admin and Member roles can approve actions in Hiro.

## Available Actions by Integration

### Okta

| Action              | Description               | Approval Level |
| ------------------- | ------------------------- | -------------- |
| Clear user sessions | Force re-authentication   | Standard       |
| Expire password     | Require password reset    | Standard       |
| Suspend user        | Lock out of all SSO apps  | Elevated       |
| Unsuspend user      | Restore access            | Elevated       |
| Deactivate user     | Permanently remove access | Critical       |
| Reset MFA factors   | Clear enrolled factors    | Standard       |

**Example proposal:**

```
PROPOSED ACTION: Suspend Okta User
─────────────────────────────────
Target: john@company.com
Reason: Account compromise detected
Evidence:
  • Login from Tor exit node (185.220.101.1)
  • Session created from previously unseen device
  • Unusual app access pattern (7 apps in 2 minutes)
Confidence: High (88%)
Impact: User will be locked out of all SSO applications

[Approve]  [Reject]  [More Info]
```

### AWS

| Action                    | Description                                                | Approval Level |
| ------------------------- | ---------------------------------------------------------- | -------------- |
| Deactivate access key     | Disable API access                                         | Standard       |
| Delete access key         | Remove API credentials                                     | Elevated       |
| Detach admin policies     | Remove AdministratorAccess, PowerUserAccess, IAMFullAccess | Elevated       |
| Terminate EC2 instance    | Stop and remove instance                                   | Critical       |
| Delete IAM user           | Remove AWS identity                                        | Critical       |
| Revoke IAM role sessions  | Invalidate assumed roles                                   | Elevated       |
| Quarantine security group | Isolate network access                                     | Elevated       |

### CrowdStrike

| Action           | Description              | Approval Level |
| ---------------- | ------------------------ | -------------- |
| Contain host     | Network-isolate endpoint | Elevated       |
| Lift containment | Restore network access   | Elevated       |
| RTR session      | Start remote response    | Elevated       |

### GitHub

| Action              | Description               | Approval Level |
| ------------------- | ------------------------- | -------------- |
| Revoke user tokens  | Invalidate access tokens  | Standard       |
| Remove collaborator | Remove repo access        | Elevated       |
| Disable user        | Block organization access | Critical       |

### Google Workspace

| Action        | Description                | Approval Level |
| ------------- | -------------------------- | -------------- |
| Sign out user | End all sessions           | Standard       |
| Suspend user  | Block Workspace access     | Elevated       |
| Wipe device   | Remote wipe managed device | Critical       |

### Slack

| Action          | Description             | Approval Level |
| --------------- | ----------------------- | -------------- |
| Deactivate user | Remove Slack access     | Elevated       |
| End sessions    | Force re-authentication | Standard       |

## Confidence Scores

Hiro provides confidence scores for proposed actions:

| Score               | Meaning                         | Recommendation                          |
| ------------------- | ------------------------------- | --------------------------------------- |
| **High (80%+)**     | Strong evidence supports action | Review briefly, approve if appropriate  |
| **Medium (50–80%)** | Reasonable evidence             | Review carefully, consider alternatives |
| **Low (below 50%)** | Limited evidence                | Investigate further before approving    |

<Info>
  Confidence scores are based on the quantity and quality of evidence, consistency across sources, and historical patterns.
</Info>

## The Approval Interface

When Hiro proposes an action, you'll see:

### Action Details

* **What**: The specific action to be taken
* **Target**: The affected user, resource, or system
* **Why**: Reason for the recommendation
* **Evidence**: Supporting findings with timestamps
* **Confidence**: How certain Hiro is
* **Impact**: What will happen when executed

### Available Responses

| Response      | Effect                                   |
| ------------- | ---------------------------------------- |
| **Approve**   | Execute the action immediately           |
| **Reject**    | Skip this action, continue investigation |
| **More Info** | Ask Hiro for additional context          |
| **Defer**     | Save for later review                    |

## Manual Instructions

When an integration doesn't support API-based remediation, Hiro provides manual instructions:

```
MANUAL ACTION REQUIRED
─────────────────────
Hiro cannot automatically suspend users in your Okta configuration.

To suspend john@company.com manually:
1. Log in to Okta Admin Console
2. Navigate to Directory > People
3. Search for "john@company.com"
4. Click the user, then Actions > Suspend
5. Confirm the suspension

Once complete, mark this action as done in Hiro.

[Mark as Complete]  [Need Help]
```

## Audit Trail

Every action (proposed, approved, rejected, executed) is logged:

```
ACTION AUDIT LOG
────────────────
2024-01-15 14:32:15 | PROPOSED | Suspend Okta user john@company.com
2024-01-15 14:32:45 | APPROVED | by admin@company.com
2024-01-15 14:32:46 | EXECUTED | Okta API returned success
2024-01-15 14:32:47 | VERIFIED | User status confirmed as SUSPENDED
```

Access the full audit trail in [**Settings > Audit Log**](https://app.hiro.is/settings/audit) or export from any Fight Mode session.

## Rollback and Recovery

Some actions support rollback:

| Action                    | Rollback         |
| ------------------------- | ---------------- |
| Suspend Okta user         | Unsuspend user   |
| Contain CrowdStrike host  | Lift containment |
| Deactivate AWS access key | Reactivate key   |
| Terminate EC2 instance    | Not reversible   |
| Delete IAM user           | Not reversible   |

<Warning>
  Destructive actions (delete, terminate) cannot be rolled back. Hiro will warn you before proposing irreversible actions.
</Warning>

## Best Practices

### Review Evidence Before Approving

Even with high confidence, take a moment to understand why Hiro is recommending an action.

### Start with Reversible Actions

When possible, prefer suspending over deleting, containing over terminating.

### Document Your Decisions

Use the notes feature to record why you approved or rejected an action.

### Verify After Execution

Hiro verifies actions automatically, but you can also confirm manually.

## Next Steps

<CardGroup cols={2}>
  <Card title="Fight Mode" icon="swords" href="/features/fight-mode">
    Use remediation in continuous threat hunting.
  </Card>

  <Card title="Integrations" icon="plug" href="/integrations/overview">
    Connect more systems for broader remediation coverage.
  </Card>
</CardGroup>
