/aport-complete — Verify Before Done

Your APort passport defines what you must deliver before a task is complete.
This skill enforces that contract. Do not mark any task done without running this first.

When to use this skill

• - Before telling a user "I'm done" or "it's complete"
• Before closing a task, ticket, or GitHub issue
• Before opening a PR and considering work finished
• Before handing off to another agent
• Any time you would otherwise call a task complete

Prerequisites

You must have an APort passport with a deliverable contract.

If you don't have one, create a passport first:

• - Web: https://aport.id
• CLI: INLINECODE0
• Agent skill: Read https://aport.id/skill and follow the instructions

You need APORT_AGENT_ID in your environment. This is returned when you create
your passport and saved to aport-passport.json.

Step 1 — Compile your completion context

Before calling the API, gather:

Step 2 — Build your attestations

Your passport defines acceptance criteria as a list of descriptions.
Each criterion maps to a criterion_id — the description slugified to lowercase
with spaces replaced by hyphens.

Example:
CODEBLOCK0

For each criterion in your passport, you must submit:
CODEBLOCK1

Evidence must be non-empty. "I believe this is met" is not evidence.
Good evidence: PR #47 at github.com/org/repo/pull/47
Good evidence: grep -r "TODO" src/ returned 0 results
Good evidence: INLINECODE22

Step 3 — Call the verify endpoint

CODEBLOCK2

Note: agent_id goes inside context, alongside the policy-specific fields.

Step 4 — Handle the response

ALLOW response

{
  "decision": {
    "decision_id": "xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx",
    "allow": true,
    "reasons": [
      { "code": "capability_verified", "message": "...", "severity": "info" }
    ],
    "created_at": "2026-03-13T09:00:00Z",
    "signature": "...",
    "policy_id": "deliverable.task.complete.v1",
    "agent_id": "ap_..."
  },
  "request_id": "..."
}

DENY response

Do not mark the task done. Check decision.reasons[].code for the deny code. Fix the issue and retry.

Step 5 — Announce completion

Once you receive allow: true, tell the user:

• - What you completed
• The decision_id (if they want the audit trail)
• Any next steps

Do not say "done" before receiving allow: true. The contract defines done, not you.

Retry behaviour

The deny_code is designed for autonomous retry. If you receive:

• - oap.summary_insufficient → rewrite your summary and retry immediately
• INLINECODE41 → add evidence strings and retry immediately
• INLINECODE42 → attempt to fix the failing tests, then retry
• INLINECODE43 → remove the blocked content, then retry
• INLINECODE44 → this requires another agent — ask the user or route to reviewer
• INLINECODE45 → attempt to resolve the criterion, then retry

Maximum retries: 3. After 3 denials on the same task, surface the issue to the user.
Do not loop indefinitely.

Links

• - Create a passport: https://aport.id (web) or npx aport-id (CLI) or https://aport.id/skill (agent)
• Your passport: https://aport.id/passport/YOURSLUG
• Verify a decision: GET https://aport.io/api/verify/decisions/get/DECISIONID
• API docs: https://aport.io/api/documentation