feat(docs): define external dogfood scorecard and stop/go criteria

[spytensor]

Problem

CoreRoom has strong internal dogfood evidence, but weak external proof. To decide whether continued optimization is worthwhile, the project needs a simple external dogfood scorecard with stop/go criteria. Without this, we may keep polishing features that do not convert real users.

Proposed solution

Add a lightweight external dogfood plan and scorecard document. It should define the smallest meaningful 2-4 week validation loop around the product thesis:

CoreRoom is valuable when it makes AI-assisted engineering work more reviewable, evidence-backed, and safe than using a single agent CLI directly.

The scorecard should specify pilot tasks, measurement fields, confusion points, and stop/go thresholds.

Alternatives considered

• Keep relying on maintainer intuition and local dogfood. Rejected: too much risk of self-confirming product work.
• Build telemetry first. Rejected: a manual scorecard is enough before adding product analytics.

Scope target

v0.x or later (deferred is fine)

Acceptance criteria

• A docs file defines the external dogfood cohort target, task shape, and pilot script.
• The scorecard captures activation, first successful issue-to-evidence run, time-to-understand value, time-to-PR-evidence, failure/confusion points, and whether the user would run it again.
• The document defines explicit continue / pivot / stop thresholds for a 2-4 week validation window.
• The document distinguishes evidence collected manually from any future automated telemetry.
• README or development docs link to the scorecard without claiming results that have not happened.

Validation

• git diff --check
• Manual review: the scorecard can be handed to an external user without extra maintainer explanation.

Out of scope

• No recruitment or private user data collection in this issue.
• No telemetry implementation.
• No marketing claims based on planned-but-not-run dogfood.

[spytensor]

Tracking umbrella: #408. This issue is part of the direct user-scoped CoreRoom control-plane polish batch opened on 2026-05-28.