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

# Governed execution

> How implementations get executed with evidence, approvals, and audit trails

Panaptico is not fully autonomous and not fully manual. It uses **governed execution** — bounded, permissioned, auditable work where humans retain control at every decision point while AI handles discovery, sequencing, and routine work.

## How tasks work

Every task in the implementation checklist follows a governed lifecycle:

<Steps>
  <Step title="Task is created with requirements">
    Each task has acceptance criteria, evidence requirements, dependencies, an owner, and optionally an approval gate
  </Step>

  <Step title="Dependencies are checked">
    A task cannot start until its upstream dependencies are completed. Blocked tasks are visible in the dependency graph.
  </Step>

  <Step title="Work is executed">
    The owner works through the task — manually, with AI assistance, or through bounded automated execution in a sandboxed environment
  </Step>

  <Step title="Evidence is attached">
    Completion requires proof: files, screenshots, test results, configuration exports, links, or verification output. Tasks cannot be marked done without evidence.
  </Step>

  <Step title="Approval is routed">
    If the task has an approval gate, it routes to the named approver. The approver can approve, request changes, or escalate.
  </Step>

  <Step title="State is recorded">
    Every state change — pending → in progress → done, approval granted, evidence attached, risk escalated — is timestamped and attributed in the audit trail.
  </Step>
</Steps>

## Evidence model

Evidence is not optional documentation. It is structured proof linked to specific tasks:

| Evidence type            | Examples                                                                       |
| ------------------------ | ------------------------------------------------------------------------------ |
| **Files**                | Configuration exports, Terraform plans, test reports, architecture screenshots |
| **Links**                | Dashboard URLs, monitoring endpoints, ticket references                        |
| **Execution results**    | Output from AI-assisted task execution in sandboxed environments               |
| **Verification results** | Automated checks confirming acceptance criteria are met                        |

Tasks can define what evidence is required before they can be completed.

## Approval chains

Approvals in Panaptico are explicit:

* Each approval gate has a **named approver** — not "someone from the team"
* Approvers see the task context, evidence, and execution results
* Three outcomes: **approved**, **changes requested**, or **escalated**
* Approval history is recorded in the audit trail
* Unresolved approvals are surfaced in the project overview as risks

## AI-assisted execution

For tasks that benefit from automation, Panaptico provides AI-assisted execution:

* **Task-scoped agents** can generate diagnostics, remediation plans, and configuration files
* Execution runs in **sandboxed environments** with explicit dependency handling
* Output is captured as evidence and linked to the task
* Agents cannot bypass approval gates or modify blueprint-wide state without explicit request

The AI assists — it does not override human judgment.

## Risk and blocker management

When work is blocked:

* Tasks can be marked as **blocked** with a reason and downstream impact
* **Risks** are tracked with severity, ownership, and resolution status
* Blocking risks surface in the project overview intelligence dashboard
* The dependency graph shows which downstream tasks are affected
* Risk resolution is recorded with evidence

## What makes this different from a ticket system

| Dimension           | Ticket system (Jira, Linear)    | Panaptico governed execution                         |
| ------------------- | ------------------------------- | ---------------------------------------------------- |
| **Structure**       | Flat or loosely grouped tickets | Phased tasks with formal dependencies                |
| **Evidence**        | Optional attachments            | Required proof linked to acceptance criteria         |
| **Approvals**       | Comment-based or external       | Named approvers with explicit routing                |
| **Context**         | Isolated per ticket             | Every task sees the full implementation graph        |
| **Audit**           | Activity log per ticket         | Full cross-task, cross-surface audit trail           |
| **Health**          | Manual status reports           | Automated scoring, trend detection, A–F grading      |
| **Post-completion** | Ticket is closed                | Task state feeds into post-implementation monitoring |

## Next steps

<CardGroup cols={2}>
  <Card title="Post-implementation" icon="chart-line" href="/docs/core-concepts/post-implementation">
    What happens after go-live
  </Card>

  <Card title="Executing tasks" icon="play" href="/docs/guides/executing-tasks">
    Step-by-step task execution guide
  </Card>
</CardGroup>