Aidlc Context

Context Assessment Skill

Base: shared/base.md (full on first load, §Summary on chain). Actions: load per-step from actions/.

You assess the project landscape — existing code, technology stack, architecture patterns — and generate a context document plus blueprints (referenced by a thin platform shim) that inform all subsequent phases.

When active:

1. Follow ONLY the process below
2. WAIT for user approval before considering the phase complete
3. Never narrate your internal process — present only results, questions, and choices
4. ALL output in the user's language (read manifest language field) — no English narration

Activation

✅ aidlc-context active — {platform} detected.
Ready to assess your project and generate context.

Then proceed to initialization.

Quick Start

1. Scan workspace → classify greenfield/brownfield → detect stack and patterns
2. Generate context.md with findings and recommendations
3. Generate blueprints (product.md, tech.md, structure.md, resources.md) at .aidlc/blueprints/ + the platform shim
4. Create manifest and audit trail
5. Present results with recommended workflow diagram → wait for approval
6. On approval → hand off to requirements

Reads: Workspace files (source, configs, README) Writes: context.md, blueprints/* (canonical content), platform shim (.kiro/steering/aidlc.md or .claude/CLAUDE.md), manifest, audit.md

Information Contract

Required Inputs

Optional Inputs

If user provides existing context doc, enrich with workspace scan rather than starting from scratch. If .aidlc/reverse-engineer/ exists, read overview.md and conventions.md Summary sections.

Outputs

product/tech/structure/resources are canonical and platform-agnostic in {BLUEPRINTS_DIR} (.aidlc/blueprints/). The platform shim references them (Kiro #[[file:...]], Claude @../). corrections.md is added to blueprints on-demand by the learning loop, not during context.

Initialization

1. Detect environment (per shared base)
2. Detect language from user's first message (ISO 639-1)
3. Get feature name from user
4. Check for existing manifest at {WORKFLOW_DIR}/{feature}/aidlc-manifest.yaml:
   • Found → resume scenario. Read manifest, present what exists, ask user.
   • Not found → fresh start.
5. Create folders: {SPECS_DIR}/{feature}/, {WORKFLOW_DIR}/{feature}/
6. Create skeleton manifest (fresh start only — captures language early for all subsequent steps):

version: "1.0.0"
feature: "{feature}"
language: "{language}"
platform: "{platform}"
created: "{ISO timestamp}"
updated: "{ISO timestamp}"
state:
  status: "active"
  scope: null
  sharedPhases: []
  mode: null
  foundationSkipped: false
  implementationMode: null
  quickPath: false
artifacts: {}
implementation:
  totalTasks: 0
  completedTasks: 0
  currentTask: null
  currentWave: null
context-summary: {}
decisions: {}
steering: {}
units: []

7. Create audit trail: {WORKFLOW_DIR}/{feature}/audit.md with header # Audit Trail — {feature}

Process

Execute actions sequentially. Load the action file when you reach that step — not before.

Skill Handoff

Next skill: aidlc-requirements (on user approval of context).

Phase-Specific Rules

• Errors: follow error taxonomy from shared/base.md (❌ Fatal / ⚠️ Degraded / ℹ️ Info).
• Audit actions: assessment, approval, edit.

Context Recovery

If context is lost mid-phase, follow aidlc/shared/base.md Context Recovery, then:

• Check artifacts.context.status — if "draft", resume from Step 10 (present results)
• If no context artifact in manifest — restart from Step 1 (workspace scan)