scope-estimate

name

scope-estimate

description

Two-module skill for the exec-db pivot. Module A captures an executive's actual workflow via canonical questions and generates USER / SYSTEM / ADMIN user stories from those answers. Module B captures PR1/PR2/PR3 technical scoping decisions and produces cost USD, clock-time weeks, and autonomous-agent terminal-time hours (plus token spend). Module A runs first by default. Honors a configurable conversation style (concise / explaining / verbose / coaching / executive-brief) read from style.md. Use when the user asks to crystallize an exec workflow, generate user stories, scope/estimate the build, or "answer the questions."

scope-estimate

Two modules, run in order:

Module A — exec workflow → user stories

1. Ask the canonical workflow questions in exec-workflow.md.
2. Persist answers to docs/exec-workflow.md.
3. Generate docs/user-stories.md (USER, SYSTEM, ADMIN lenses) using the rules in user-stories.md.

Module B — scope estimate

1. Ask the technical scoping questions in questions.md.
2. Persist answers to docs/scope-answers.md.
3. Compute cost USD, clock-time weeks, and autonomous-agent terminal-time hours + token spend using estimation.md.

Step 0 — load the style setting

Before doing anything else, read style.md and apply the active style to every piece of assistant-generated prose for the rest of the run (questions, summaries, narrative inside generated docs). The user may override with phrases like "use coaching style" or "answer in executive-brief" — per-invocation overrides win over the file. Numeric estimates, IDs, headings, and table formulas are unaffected by style.

When to invoke

• "crystallize my workflow" / "interview me" / "run the exec questions"
• "generate user stories" / "give me USER/SYSTEM/ADMIN stories"
• "answer the scope questions" / "estimate this build"
• "scope/budget the exec-db pivot"
• user pastes the long question list and asks for estimates

If the user names only one module, run only that module. Otherwise default to A then B, and confirm before starting B if A produced any unanswered W-ids.

Module A workflow

1. Locate or create docs/exec-workflow.md from exec-workflow-template.md. Preserve any existing answers.
2. Collect answers to W1–W10 in batches of 6–10. Quote the exec verbatim where possible — these are inputs to story generation. Save each batch immediately.
3. Generate docs/user-stories.md from user-stories-template.md, applying every rule in user-stories.md:
   • Three lenses (USER / SYSTEM / ADMIN), no duplicates.
   • Stable IDs US-###, SY-###, AD-###.
   • Each story tagged with source W-id, t-shirt size, must/should/could, and links to S-ids from the technical questionnaire.
   • Acceptance criteria for every M/L story.
   • List unanswered W-ids under ## Unblockers.
4. Show a short summary in the active style: counts per lens, any cross-cutting invariants discovered, and the top 3 stories the exec should sanity-check.

Module B workflow

1. Locate or create docs/scope-answers.md from template.md.
2. Collect answers to S1–S10. If the user says "use defaults" / "MVP defaults", apply defaults.md and tag each answer (default). If docs/user-stories.md exists, prefer answers consistent with the priority of stories already marked must.
3. Compute estimates using the deterministic line items in estimation.md. Do not guess totals — sum the rows.
4. Write the Estimates block at the bottom of docs/scope-answers.md:
   • Per-PR breakdown (PR1 foundation, PR2 Google + autodraft, PR3 PM digests, vision-check CLI).
   • Cost low/high USD.
   • Clock-time low/high weeks (1 senior full-stack dev at 30 productive hr/wk unless user says otherwise).
   • Terminal-time low/high agent wall-clock hours, human-review hours (≈10%), and token spend at $3.50/agent-hour cached ($11/agent-hour uncached).
   • Top 3 cost drivers and top 3 deferrable items.
5. Show a short summary in the active style: the three numbers, the assumptions that drove them, and the path to the file.

Files in this skill

• style.md — active conversation style + style definitions.
• exec-workflow.md — canonical exec workflow questionnaire (W-ids).
• exec-workflow-template.md — skeleton for docs/exec-workflow.md.
• user-stories.md — rules for generating user stories.
• user-stories-template.md — skeleton for docs/user-stories.md.
• questions.md — technical scoping questionnaire (S-ids).
• template.md — skeleton for docs/scope-answers.md.
• defaults.md — MVP defaults for every S-id.
• estimation.md — line items, rates, token model, sanity bands.

Hard rules

• Always read style.md first and apply the active style to all assistant-generated prose.
• Always write to the canonical paths (docs/exec-workflow.md, docs/user-stories.md, docs/scope-answers.md) unless the user passes a different path.
• Never invent answers the user did not give and did not delegate to defaults. Mark unknown items _unanswered_ and exclude them from estimates and from story generation; flag them as blocking.
• Never alter the wording of W- or S- questions to fit a style.
• Stories must be testable. No "user has a great experience" stories.
• Estimates are ranges, not point numbers. Low = best case with defaults honored; High = realistic case with ~30% scope creep.
• Terminal-time and clock-time are independent estimates — do not derive one from the other.
• Do not commit the generated docs automatically. Ask first.