estimator

Project estimation tool with interactive workflow. Breaks down projects into functional components, calculates cost ranges (optimistic/realistic/pessimistic) with auto-detected risk coefficients. Gene

SKILL.md

GitHub

Estimator — Project Estimation Tool

Overview

Estimator transforms functional requirements into structured, argumented cost estimates through an interactive 4-phase workflow with mandatory checkpoints. Outputs client-ready Markdown documents consumable by Propositor.

Quick Decision Tree

┌─────────────────────────────────────────────────────────────────┐
│                    What type of estimation?                      │
└─────────────────────────────────────────────────────────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        ▼                   ▼                   ▼
┌───────────────┐   ┌───────────────┐   ┌───────────────┐
│  NEW PROJECT  │   │  EVOLUTION    │   │  AUDIT/TMA    │
│  (dev/refonte)│   │  (features)   │   │  (review)     │
└───────┬───────┘   └───────┬───────┘   └───────┬───────┘
        │                   │                   │
        ▼                   ▼                   ▼
   Full workflow       Partial lots        Specific template
   (all 12 lots)       (impacted only)     (no recette for audit)

Main Workflow (4 Phases)

Phase 1: Qualification

Goal: Understand context and calibrate estimation

  1. Analyze input (brief, brainstorm output, specs)
  2. Ask clarification questions (≤3 if brief is clear)
  3. Auto-detect: project type, granularity, coefficients

📍 Checkpoint 1: Present understanding + detected parameters for validation

→ See workflow-details.md for checkpoint format

Phase 2: Functional Breakdown

Goal: Identify ALL features (explicit AND implicit)

  1. Extract features from brief
  2. Identify implicit features (auth, logs, admin...)
  3. Assign unique IDs (FCT-001, FCT-002...)
  4. Prioritize: MVP / Should / Could
  5. Provide AI suggestions

📍 Checkpoint 2: Present feature table + suggestions for validation

Phase 3: Task Evaluation

Goal: Estimate each task with ranges

  1. Generate task tables per lot
  2. Calculate: JH_Low × 0.8 | JH_Mid × coeff | JH_High × 1.3
  3. Apply automatic recette (15% for dev, varies by type)

📍 Checkpoint 3: Present detailed estimation for validation

→ See lots-templates.md for lot structures

Phase 4: Valorization & Synthesis

Goal: Convert JH to budget, document assumptions

  1. Apply TJM (default 450€)
  2. Generate budget scenarios (Light/Low/Mid/High)
  3. Document technical registry (assumptions, risks, stack choices)

📍 Final Checkpoint: Complete document ready for export or Propositor

Granularity Auto-Detection

Project Size Granularity Lots Structure
< 30 JH Macro (±30%) 4 merged lots
30-200 JH Standard (±20%) 12 standard lots
> 200 JH Detailed (±10%) 12 lots + Back/Front sub-modules

Coefficient Auto-Detection

Client Type Specs Clarity coeff_effort coeff_risk
Known Clear 0.85 1.05
Known Partial 0.90 1.10
New Clear 0.90 1.10
New Unclear 0.95 1.20

→ See coefficients.md for detailed formulas

Tech Stack Preferences

Domain Priority 1 Priority 2 Fallback
Backend Symfony 7 Django Spring Boot
Frontend React 18 Vue 3 Angular
Database PostgreSQL MySQL MongoDB
Infra Docker PaaS VM
Mobile React Native Flutter Native

Commands Reference

During Session

Command Action
valider Confirm checkpoint, proceed to next phase
ajouter [item] Add feature or task
modifier [ID] Edit existing element
supprimer [ID] Remove element
question [topic] Ask clarification before validating
ajuster-jh [ID] [value] Manually adjust JH
recalculer Recalculate after modifications
détailler [lot] Show lot details
exporter Generate final document
propositor Chain to Propositor skill

Launch Flags

Flag Effect Default
--macro Force macro granularity (±30%) Auto
--standard Force standard granularity (±20%) ✅
--detailed Force detailed granularity (±10%) Auto if >200 JH
--tjm [amount] Set specific TJM 450€
--type [dev/refonte/tma/audit] Force project type Auto-detected
--coeff [0.6-1.0] Override effort coefficient Auto
--no-suggestions Disable AI suggestions Enabled

Output Format

Estimator produces structured Markdown with parseable tags:

<!-- ESTIMATOR_DATA_START -->
| Lot | JH Low | JH Mid | JH High |
|-----|--------|--------|---------|
...
<!-- ESTIMATOR_DATA_END -->

<!-- ESTIMATOR_BUDGET_START -->
| Scenario | JH | Amount HT |
|----------|-----|-----------|
...
<!-- ESTIMATOR_BUDGET_END -->

→ See output-format.md for complete template

Critical Rules

  1. Mandatory checkpoints — No one-shot generation
  2. Proactive AI suggestions — Implicit features, risks, alternatives
  3. Ask when uncertain — Never guess, request clarification
  4. FCT-xxx traceability — Each task references a feature
  5. Complete technical registry — Document all assumptions
  6. User's language — Output in input language
  7. Parseable tags — For Propositor integration

Synergies

Inputs From

  • brainstormer → Synthesis report, decisions
  • code-promptor → Technical brief, constraints
  • resumator → Meeting notes, requirements

Outputs To

  • propositor → Primary — Full estimation for commercial proposal
  • planificator (future) → JH per lot for detailed planning
  • tracker (future) → Initial estimate as tracking reference

Limitations

This skill does NOT handle:

  • Invoicing or accounting
  • Contracts or legal aspects
  • Estimates without technical scope
  • Negotiation (see negociator)
  • Detailed planning (see planificator)

Version History

Version Date Changes
1.0.0 2025-12-15 Initial release

Current: v1.0.0

Owner

  • Author: Édouard
  • Contact: Via Claude.ai