desk-flip — Skill Interview Log
Assumptions Checkpoint
| # | Assumption | Tag | Status |
|---|---|---|---|
| 1 | Target is existing projects that are feature-complete / buttoned-down and stuck | [from request] | Confirmed |
| 2 | Core problem is infrastructure built before prototype/taste alignment, making UAT hard | [from lessons] | Confirmed |
| 3 | "Start from scratch" means restarting the workflow with existing code as reference | [inferred] | Corrected — user means full reset, literally starting the project over |
| 4 | New workflow requires specs/, tasks/roadmap.md, etc. — stuck projects lack these | [from codebase] | Confirmed (implicit) |
| 5 | Skill should be repo-managed and global | [inferred] | Confirmed |
| 6 | Skill should assess whether to migrate vs. restart | [inferred] | Corrected — always desk-flip, no assessment needed |
Interview Questions and Decisions
Turn 1: Core Shape
Q1: Is "start from scratch" workflow-restart or full reset?
- Options: Restart workflow only / Full reset / Skill decides per-project
- Recommendation: Restart workflow only
- User chose: Full reset
- Impact: The skill doesn't preserve the old codebase as the working tree. Old code stays in its repo, new project is a separate repo.
Q2: What triggers this skill?
- Options: Manual only / Manual + recommended / Auto-detection first
- Recommendation: Manual + recommended
- User chose: Manual invocation only
- Impact: No detection hooks in other skills needed. Simpler contract.
Q3: Single skill or assess-then-route?
- Options: Single skill / Assess then route / Assess + scaffold, route for content
- User response: "It depends on the strategy of unstucking a project"
- Impact: Led to follow-up questions about execution depth.
Turn 2: Desk-Flip Mechanics
Q4: What happens to old code?
- Options: Git branch archive / Directory archive / Separate repo / Just delete
- User chose: Separate repo
- Impact: Old repo is untouched. New project starts in a fresh repo.
Q5: Assessment signals for migrate vs. desk-flip?
- Options: Missing workflow artifacts / Alignment failure / Both + user judgment / Always desk-flip
- User chose: Always desk-flip
- Impact: Eliminates the assessment/decision branch entirely. Skill is simpler — always autopsy + extract + route.
Q6: Execution depth?
- Options: Full execution / Assess + archive, then route / Assess only, route everything
- User chose: Assess + archive, then route
- Impact: Skill does the autopsy and extraction, then routes to
/bootstrap-repo. Does not execute the fresh start itself.
Turn 3: Extraction and Naming
Q7: What to extract from the old project?
- Options: Lessons + domain knowledge / Lessons + specs + assets / Full inventory
- User chose: Lessons + specs + assets
- Impact: Extraction includes lessons, valid specs/designs, and non-code assets. No code, dependencies, or infrastructure carried forward.
Q8: Skill name?
- Options: desk-flip / fresh-start / project-reset
- User chose: desk-flip
Coverage Checkpoint
Presented full coverage table. User confirmed "Looks good, write the brief" with no corrections.
Deviations from Initial Request
- User initially asked whether
/hygienewas adequate — determined it is not (hygiene is structural linting, not workflow migration) - User initially framed this as "migrate" — evolved to "always desk-flip" through the interview
- Initial assumption of incremental migration as an option was eliminated — the skill is simpler and more opinionated as a result