← ClaudeAtlas

jtbd-contextlisted

Stage 01 of JTBD pipeline. Reads project materials and writes factual context separating facts, hypotheses, unknowns. Use when user asks to analyze a project, product, brief, offer, UX, or content through JTBD. Entry point — start here.
stepanenkoviktor0110-boop/ai-dev-methodology · ★ 1 · AI & Automation · score 57
Install: claude install-skill stepanenkoviktor0110-boop/ai-dev-methodology
# JTBD Context (Stage 01 of 4) ## Pipeline position Stage 01 of 4: **jtbd-context** → jtbd-demand-diagnosis → jtbd-job-model → jtbd-project-decisions. State passes through markdown files in `jtbd-analysis/` in the target project root. ## Purpose Build a factual base for the JTBD analysis. Do **not** interpret motivation, demand, or jobs here — that is stages 02-04. ## Working folder Create `jtbd-analysis/` in the target project root if it does not exist. Write the output to `jtbd-analysis/01_context.md`. ## Steps 1. **Ask the user which materials to read.** Do not auto-scan the project. List the sources you will use (brief, website, README, landing, mocks, interview notes, analytics). 2. **Extract project facts** without motivation interpretation: - what the project is (1-2 sentences), - what it does (concrete actions, features, deliverables), - apparent users, apparent buyers, - claimed problem and promised result (as stated in materials), - constraints (technical, business, regulatory, time, budget). 3. **Tag every statement** as fact / hypothesis / unknown. Facts reference their source. Hypotheses include reasoning. Unknowns name the gap. 4. **Rate evidence quality** as High / Medium / Low with one-line justification. 5. **Self-validate** against `references/quality-checklist.md` (Universal rules + Stage 01 section). Fix failures. 6. **If the same checklist item fails three times** — stop silent fixing and ask the user the relevant question from `ref