Spec Review

The most common Profer workflow. Your agent writes a spec, you get it reviewed before building.

The prompt

Tell your agent:

"Write a technical spec for the auth migration. Include the database schema changes, API endpoints, and migration strategy. Publish it for review with approval and concerns questions."

What happens

1. Agent generates HTML with the full spec
2. Agent calls publish() with questions:

{
  "html": "<h1>Auth Migration Spec</h1>...",
  "title": "Auth Migration — Technical Spec",
  "questions": [
    { "id": "approval", "type": "approve", "text": "Ready to implement?" },
    { "id": "approach", "type": "choice", "text": "Which migration strategy?", "options": ["Big bang", "Gradual rollout", "Feature flag"] },
    { "id": "concerns", "type": "text", "text": "Any concerns or blockers?" }
  ]
}

3. You get back a URL — paste it in Slack or email
4. Tech lead clicks, reads, picks "Gradual rollout", approves, notes a concern about backward compatibility

Reading feedback

Ask your agent:

"Check the feedback on the auth migration spec"

Agent calls get() and reads:

{
  "status": "approved",
  "feedback": [{
    "reviewer": "Sarah",
    "answers": {
      "approval": "yes",
      "approach": "Gradual rollout",
      "concerns": "Need to maintain backward compat with v1 tokens for 30 days"
    }
  }]
}

Agent now knows: approved, gradual rollout, 30-day backward compat requirement. It can start implementing with this context.

Why this works

• 3 minutes from "write a spec" to reviewed spec
• Reviewer answers structured questions, not free-form comments
• Agent gets typed data back, not prose to parse
• No Google Doc needed. No Jira ticket. No meeting.