Code-Promptor — Transcript to Development Brief

Overview

Code-Promptor transforms raw audio transcriptions or unstructured text into clean, professional development briefs. Version 2.1 adds session mode for batch processing, multi-task detection, and direct Notion integration.

Core Principle: Faithful extraction, never invention. Generate actionable briefs with intelligent subtask suggestions.

Language: Output matches input language. Mixed input defaults to French structure with English technical terms preserved.

Decision Tree

┌─────────────────────────────────────────────────────────────────┐
│                    How is Promptor activated?                    │
└─────────────────────────────────────────────────────────────────┘
                            │
            ┌───────────────┴───────────────┐
            ▼                               ▼
┌───────────────────────┐       ┌───────────────────────┐
│   SINGLE MODE         │       │   SESSION MODE        │
│   "promptor" + input  │       │   "promptor session"  │
└───────────┬───────────┘       └───────────┬───────────┘
            │                               │
            ▼                               ▼
    Process one dictation           Init session + project
    Full workflow                   → Each input isolated
                                    → Multi-task detection
                                    → Direct Notion export

Session Mode

Activation Triggers

• promptor session / session promptor
• mode série promptor / batch promptor

Session Workflow

1. INIT: User activates session
   → Ask for Notion project (optional)
   → Confirm session active

2. DICTATION RECEIVED:
   → Clean voice artifacts
   → Detect mono vs multi-task
   → If multi: show checkpoint for validation
   → Generate brief(s)
   → Export to Notion
   → Reset context for next dictation

3. END: "fin session" or topic change
   → Show summary with all created tasks

Session Commands

Command	Action
projet [nom]	Change project mid-session
status	Show session state
fin session	End and show summary

→ See Session Init Message

Multi-Task Detection

Mode: AGGRESSIVE

The skill tends toward detecting multiple tasks. User can merge if needed.

Detection Algorithm

1. Clean dictation (keep rupture markers)
2. Segment on explicit/implicit markers
3. Score each segment for independence
4. Decide: ≥2 segments with score ≥40 → MULTI-TASK

Rupture Markers

Type	Markers	Points
Explicit	"aussi", "et puis", "autre chose", "ah et", "sinon"	+30
Implicit	Domain change, subject change	+15-25

Independence Score

SCORE = SUBJECT_DIFF×25 + ACTION_DIFF×20 + DOMAIN_DIFF×25 + MARKERS

→ See Multi-Task Detection for full algorithm

Checkpoint Validation

When multi-task detected, show validation checkpoint:

📋 **N tâches détectées**

┌───┬─────────────────────────┬──────────┬────────────┬───────┐
│ # │ Titre suggéré           │ Type     │ Complexité │ Temps │
├───┼─────────────────────────┼──────────┼────────────┼───────┤
│ 1 │ [Title]                 │ [Type]   │ [Level]    │ [Est] │
└───┴─────────────────────────┴──────────┴────────────┴───────┘

📝 Segments extraits: [show source segments]

📖 Commandes:
   ok          Générer tous les briefs
   ok 1,2      Sélection partielle
   merge 1,2   Fusionner
   edit N "x"  Modifier titre
   drop N      Supprimer
   split N     Découper
   reanalyze   Relancer détection

→ See Checkpoint Format for full template

Complexity Levels

Level	Criteria	Time	Plan
Quick fix	<50 words, corrective verb	1h	❌
Standard	50-200 words, clear scope	4h	✅
Major	>200 words, multi-component	8h	✅ Detailed

→ See Output Format for templates per level

Brief Structure

Common Elements (all levels)

# [Action Verb] + [Object] — Notion-ready title

📦 **[Complexity]** | ⏱️ [Time] | 🎯 Confidence: [Level]

## Objectif
[2-4 sentences]

## Description  
[Context and functioning]

## Exigences fonctionnelles
- [FR list]

Standard/Major additions

## Plan d'implémentation

1. **[Phase 1]**
   - [ ] Subtask auto-generated
   - [ ] Subtask auto-generated

2. **[Phase 2]**
   - [ ] Subtask auto-generated

→ See Subtask Templates for generation rules

Notion Integration

Properties Filled

Property	Source
Nom	Brief title
Description	Brief content
Type	Auto-detected
Temps estimé	Based on complexity
Projet	Session init

Type Mapping

Detection	Notion Type
Bug, fix, corriger	Bloquant
Feature, créer	Evolution
Refacto, generic	Tache
Backend specific	Backend
Frontend specific	Frontend

→ See Type Mapping for full rules

Error Handling

If Notion API fails:

1. Display complete brief in text
2. Show error message
3. Offer retry option

Additional Commands

Command	Usage
ref [n]	Reference task n (creates dependency)
--no-plan	Disable subtask generation
--no-estimate	Disable time estimation

Critical Rules

1. Never ask questions — Produce brief with available info
2. Never invent requirements — Mark absent if not mentioned
3. Never reference source — Brief is self-contained
4. Each dictation = isolated context — No pollution between inputs
5. Later wins — Last stated version overrides earlier

Knowledge Base

References

• Output Format — 3 brief templates
• Multi-Task Detection — Detection algorithm
• Subtask Templates — Auto-generation rules
• Type Mapping — Notion type mapping
• Processing Rules — Extraction methodology
• Voice Cleaning — Dictation cleanup

Config

• Notion IDs — Database configuration
• Projects Cache — Project name resolution

Templates

• Checkpoint Format — Validation display
• Brief Quick Fix
• Brief Standard
• Brief Major

Limitations

This skill does NOT:

• Ask clarifying questions (produces with available info)
• Execute development tasks (pre-processor only)
• Modify existing Notion tasks
• Handle bulk imports from files
• Work on already-structured documents

Version History

Version	Date	Changes
2.0.0	2025-12	Initial Claude skill (from GPT migration)
2.1.0	2025-12-31	Session mode, multi-task detection, Notion integration, subtask generation

Current: v2.1.0

Owner

• Author: Édouard
• Contact: Via Claude.ai