Audit a draft spec for missing edge cases

The bugs that ship aren't the ones you wrote down. They're the ones you didn't think to write down. This use case is about getting a second pair of eyes on a draft spec — eyes that have read thousands of similar specs and know what tends to be missing.

The input

Any draft spec. Could be a Google Doc, a Notion page, a Linear ticket description, or a markdown file. Even a few paragraphs is enough.

# Goal
Add a "Forgot password" flow to the login screen.

# Outcomes
- User can request a reset email
- User can set a new password from the email link
- User is logged in after reset

The flow

1. Create a new intent. From your product, click New intent. Paste your draft into the conversation panel.

2. Let the AI interrogate it. Pathmode's conversation runs Socratic by default — it doesn't just rewrite your spec, it pushes back. Expect questions like:

• "What happens if the reset link is clicked twice?"
• "You haven't said how long the link stays valid. What's your assumption?"
• "Is the user's session on other devices invalidated after reset?"

3. Watch the spec crystallize. As you answer, edge cases get added to the Spec panel on the right. Each one shows up as an InlineSpecCard you can undo within 10 seconds if it's not quite right.

4. Check the ThinkingQuality score. Four dimensions: specificity, testability, coherence, tension. A high score on tension means the spec actually grapples with conflicts — not just lists happy-path outcomes.

The output

A draft spec that's been pressure-tested. Typical session adds:

• 3–6 new edge cases the original draft didn't mention
• Implicit assumptions made explicit as constraints
• Verification criteria that distinguish "feature works" from "feature ships safely"
• Open questions flagged rather than papered over

Try it yourself

1. Sign in to Pathmode
2. Open a product
3. Click New intent
4. Paste any draft spec into the conversation
5. Reply to the AI's questions — that's where the value lives

Related

• Playbook: Why Specs Fail
• Playbook: The Anatomy of an Agent-Ready Spec
• Use case: Go from a vague founder-ask to a testable goal