Specs Quality - Context Improvement

Maintains the quality and consistency of specification context by synchronizing Knowledge Graph, Tasks, and Codebase.

Overview

This command solves three main problems in the specification workflow:

1. Inconsistent Technical Context: Tasks lose technical context or don't reflect actual patterns used in the codebase
2. Specs-Tasks Misalignment: User request, specification, and tasks are not aligned
3. Obsolete Knowledge Graph: The knowledge-graph.json is not updated after implementations

Workflow Position

Idea → Specs → Tasks → Implementation → Specs Quality Update (this)
                ↑         ↓              ↓
                └─────────────────────────────────────┘
                    Continuously sync context

Usage

# Basic usage - update context for a spec folder
/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/

# Update KG only (used by spec-to-tasks after codebase analysis)
/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/ --update-kg-only

# Update after task completion (used by task-implementation)
/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/ --task=TASK-001

# Dry run - show what would be changed without making changes
/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/ --dry-run

Arguments

Argument	Required	Description
spec-folder	No	Path to spec folder (default: current working directory)
--update-kg-only	No	Only update Knowledge Graph, skip task enrichment
--task	No	Update context after specific task completion
--dry-run	No	Show planned changes without executing them

Core Principles

• Incremental updates: Only update what has changed, don't rewrite everything
• Bidirectional sync: Both KG → Tasks and Tasks → KG alignment
• Traceability: All changes are logged and reported
• Non-destructive: Preserve manual edits and annotations
• Codebase-first: Actual implementation is source of truth

---

Phase 1: Discovery

Goal: Identify spec folder and gather current context state

Actions:

1. Create todo list with all phases
2. Parse arguments:
   • Extract spec-folder path
   • Detect flags: --update-kg-only, --task=TASK-XXX, --dry-run
3. Determine spec folder:
   • If argument provided: use it
   • If no argument: detect from current working directory
   • Validate path contains spec files
4. Read current state:
   • Check if knowledge-graph.json exists
   • List all task files in tasks/ directory
   • Read user-request.md if exists
5. Detect language from existing tasks or files

---

Phase 2: Gap Analysis

Goal: Identify discrepancies between KG, tasks, and actual codebase

Actions:

1. Knowledge Graph vs Codebase:

   • For each component in KG: check if file actually exists
   • For each API in KG: check if endpoint exists
   • Find new files not documented in KG

2. Tasks vs Knowledge Graph:

   • Check if task technical context matches KG patterns
   • Identify tasks referencing non-existent components
   • Find tasks missing expected technical details

3. Requirements Traceability:

   • Compare user-request.md with task descriptions
   • Identify requirements mentioned but not in tasks
   • Find tasks without clear requirement origin

4. Generate gap report:

   ## Gap Analysis Report
   
   ### Missing in KG
   - NewFile.java (discovered in codebase)
   
   ### Outdated in KG
   - OldComponent.java (file removed)
   
   ### Tasks Needing Update
   - TASK-003: References outdated pattern
   - TASK-007: Missing technical context
   
   ### Orphaned Requirements
   - User request mentions "X" but no task covers it
   

---

Phase 3: Extraction

Goal: Extract structured information from implemented code

Actions:

1. If --task=TASK-XXX specified:

   • Read task file to identify implemented files
   • Extract from provides section or Files to Create
   • Validate files exist in codebase

2. If no specific task:

   • Scan codebase for files matching spec patterns
   • Use Glob to find recently modified files

3. Extract symbols from files:

   • For each file, use appropriate extraction method based on language
   • Java: Grep for class|interface|enum declarations
   • TypeScript: Grep for class|interface|function|const declarations
   • Python: Grep for class|def declarations

4. Classify by type:

   • Infer from directory structure: /domain/entity/ → entity
   • Infer from annotations: @RestController → controller
   • Default to generic type if unclear

5. Build provides objects:

   {
     "task_id": "TASK-001",
     "file": "src/main/java/.../Search.java",
     "symbols": ["Search", "SearchStatus"],
     "type": "entity",
     "implemented_at": "2026-03-15T10:30:00Z"
   }
   

---

Phase 4: Knowledge Graph Update

Goal: Update knowledge-graph.json with new findings

Actions:

1. Load existing KG (or create new structure):

   {
     "metadata": {
       "spec_id": "...",
       "feature_name": "...",
       "created_at": "...",
       "updated_at": "2026-03-15T...",
       "version": "1.0",
       "analysis_sources": [...]
     },
     "codebase_context": {...},
     "patterns": {...},
     "components": {...},
     "provides": [...],
     "apis": {...},
     "integration_points": [...]
   }
   

2. Update provides array:

   • Add new provides from Phase 3
   • Check for duplicates by task_id + file
   • Update implemented_at for existing entries
   • Mark entries as verified

3. Update components (if new discovered):

   • Add to appropriate category (controllers, services, etc.)
   • Preserve existing entries

4. Update APIs (if new endpoints discovered):

   • Scan for REST annotations
   • Add to internal/external as appropriate

5. Update metadata:

   • Set updated_at to current ISO timestamp
   • Add entry to analysis_sources: {"agent": "spec-quality", "timestamp": "..."}

6. Write updated KG:

   • If --dry-run: Show diff instead of writing
   • Otherwise: Write to docs/specs/[ID]/knowledge-graph.json

---

Phase 5: Task Enrichment

Goal: Update task files with improved technical context

Skip if: --update-kg-only flag is set

Actions:

1. Identify tasks needing update (from Phase 2 gap analysis)

2. For each task file:

   • Read current content
   • Parse YAML frontmatter
   • Identify "Technical Context" section

3. Enrich technical context with KG data:

   • Add relevant patterns from KG
   • Reference existing components to integrate with
   • Document APIs to use/extend
   • Note conventions to follow

4. Update provides section (if task was implemented):

   • Add or update provides: array in frontmatter
   • Include file paths, symbols, types

5. Preserve manual content:

   • Don't overwrite custom notes
   • Preserve acceptance criteria
   • Keep manual edits to descriptions

6. Write updated task file:

   • If --dry-run: Show proposed changes
   • Otherwise: Write back to task file

Phase 5.5: Spec Document Drift Detection

Goal: Detect drift between specification and implemented reality

Actions:

1. Resolve and read the spec file (YYYY-MM-DD--feature-name.md preferred, legacy *-specs.md supported):

   • Extract all acceptance criteria from the spec
   • List all user stories and requirements

2. Read decision-log.md if exists:

   • Extract all DEC entries with implementation deviations
   • Identify scope changes, additions, removals

3. Read completed tasks:

   • Scan tasks/ directory for completed tasks (status: completed)
   • Extract acceptance criteria from each completed task
   • Identify what was actually implemented

4. Compare and identify drift:

   • Criteria in spec but NOT implemented → Unmet requirements
   • Features implemented but NOT in spec → Scope expansion
   • Modified implementations → Requirement refinement
   • Documented deviations in decision-log → Explained drift

5. Generate drift report:

   • If drift detected:
     • Add to report: "Spec Drift Detected: N criteria diverged from implementation. Consider running /devkit.spec-sync [spec-folder] to update specification."
     • List specific drift items with references to DEC entries
   • If no drift:
     • Log: "Spec and implementation are aligned, no drift detected"

---

Phase 6: Report Generation

Goal: Generate summary of all changes made

Actions:

1. Generate change summary:

   ## Specs Quality Update Summary
   
   **Spec**: docs/specs/[ID]/
   **Timestamp**: [ISO timestamp]
   **Mode**: [full|kg-only|task=TASK-XXX|dry-run]
   
   ### Knowledge Graph Updates
   - Added 3 new provides entries
   - Updated 2 existing entries
   - Verified 15 components
   
   ### Task Updates
   - Enriched TASK-001 technical context
   - Enriched TASK-003 technical context
   - Added provides to TASK-007
   
   ### Gap Analysis Results
   - All requirements covered
   - Technical context synchronized
   
   ### Files Modified
   - knowledge-graph.json
   - tasks/TASK-001.md
   - tasks/TASK-003.md
   

2. If dry-run mode: Include detailed diff preview

3. Present report to user

---

Integration Points

In spec-to-tasks Command

Add to spec-to-tasks Phase 3.5 (after Codebase Analysis):

## Phase 3.5: Update Knowledge Graph

After codebase analysis completes, automatically update KG:

/developer-kit:devkit.spec-quality [spec-folder] --update-kg-only

This persists agent discoveries into knowledge-graph.json for future reuse.

In task-implementation Command

Add to task-implementation, after T-6 (Task Completion):

## T-6.5: Update Specs Quality

After task completion and verification, update spec context:

/developer-kit:devkit.spec-quality [spec-folder] --task=[TASK-ID]

This updates:
- Knowledge Graph with new provides entries
- Task file with implementation details
- Technical context for dependent tasks

---

Error Handling

Spec Folder Not Found

• Behavior: Error and ask for correct path
• Message: "Spec folder not found at [path]. Please provide valid path."

Knowledge Graph Corrupted

• Behavior: Backup corrupted file, create new structure
• Message: "KG corrupted, backed up to [path].kg.backup, creating new."

Task File Not Found (with --task)

• Behavior: Warning, continue with KG update only
• Message: "Task TASK-XXX not found, updating KG only."

File Write Failure

• Behavior: Log error, report in summary
• Message: "Failed to write [file]: [error]"

---

Best Practices

When to Run Manually

• After completing several tasks to sync context
• Before starting new implementation to verify context
• After significant refactoring
• When context seems stale or inconsistent

Automation Frequency

• After each task: Recommended for accurate tracking
• Batch updates: Acceptable for minor tasks
• Before releases: Always run to ensure context is current

KG Freshness Indicators

• < 7 days: Fresh, use cached analysis
• 7-30 days: Getting stale, consider refresh
• > 30 days: Old, recommend full refresh

---

Examples

Example 1: Full Context Update

/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/

Output:

Analyzing spec folder: docs/specs/001-hotel-search-aggregation/
Found knowledge-graph.json (3 days old)
Found 10 task files

Gap Analysis:
- 2 new components discovered
- 1 task needs technical context update

Updating Knowledge Graph...
✓ Added 2 provides entries
✓ Updated metadata

Updating Tasks...
✓ Enriched TASK-003 technical context

Summary:
- KG updated: 3 days ago → now
- Tasks enriched: 1
- Files modified: 2

Example 2: Dry Run

/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/ --dry-run

Shows what would change without applying changes.

Example 3: After Task Implementation

/developer-kit:devkit.spec-quality docs/specs/001-hotel-search-aggregation/ --task=TASK-001

Updates KG with provides from TASK-001 and enriches related tasks.

---

Todo Management

Maintain todo list:

[ ] Phase 1: Discovery
[ ] Phase 2: Gap Analysis
[ ] Phase 3: Extraction
[ ] Phase 4: Knowledge Graph Update
[ ] Phase 5: Task Enrichment (skipped if --update-kg-only)
[ ] Phase 6: Report Generation

---

Notes

• This command is designed to be idempotent: running multiple times produces same result
• Manual edits to task files are preserved whenever possible
• The Knowledge Graph is the single source of truth for technical context
• When in doubt, the actual codebase is the final authority