Closing a Feature: What Happens to Your Work

Your feature is approved. The agent's work is complete. Now what happens when you close it?

Closing a feature is not a deletion. It's a structured return of knowledge — the decisions, code, and context generated during the feature are captured and indexed back into your workspace.

The 5-step close pipeline

When you click Close, Fabriq runs this pipeline. It's idempotent and resumable — if a step fails, you can retry from where it left off.

┌─ Promote ─→ Link ─→ Index ─→ Summary ─→ Audit ─┐

Step 1: Promote

Key docs from the feature are copied from features/<name>/docs/ to the workspace knowledge area:

What gets promoted	Source	Destination
Business design	features/<f>/docs/03-business-design.md	docs/features/<f>/business-design.md
Technical design	features/<f>/docs/04-technical-design.md	docs/features/<f>/technical-design.md
Validation report	features/<f>/reports/validation-report.md	docs/features/<f>/validation-report.md
Close summary	features/<f>/reports/close-summary.md	docs/features/<f>/close-summary.md

Docs that are not promoted remain in the feature directory (available in the feature history archive).

Step 2: Link

A feature-history entry is created at:

docs/feature-history/<feature-name>/index.md

This entry contains:

• Feature name and description
• Author and date
• Links to promoted docs and reports
• Key decisions made during the feature
• Link to the audit trail entries

Step 3: Index

Any cognitive assets generated during the feature — behaviors, rules, state machine fragments discovered by CDR — are indexed into the workspace knowledge graph. This makes them searchable and navigable for future features and agents.

Step 4: Summary

A close summary is written to reports/close-summary.md:

# Close Summary: add-dark-mode-toggle

## Deliverables
- DarkModeToggle component (src/components/DarkModeToggle.tsx)
- Settings page integration (/pages/Settings.tsx, +8 lines)
- Theme CSS variables (/styles/theme.css, +15 lines)

## Test Results
- 4/5 scenarios passed
- 1 known issue: animation flicker on page load (minor, tracked)

## Must-fix Items
- None — all items resolved before approval

## Key Decisions
- Used CSS custom properties for theming (not CSS modules)
- Stored preference in localStorage (not cookies)
- Reused existing Button component

Step 5: Audit

A final audit log entry is written to .dapei/runtime/audit/capability.log:

{
  "schema_version": 1,
  "timestamp": "2026-07-20T15:00:00Z",
  "capability": "feature.close",
  "version": "0.1.0",
  "ok": true,
  "duration": 250,
  "metadata": {
    "feature": "add-dark-mode-toggle",
    "dimension": "feature",
    "promoted": 4,
    "indexed": true
  }
}

What persists after close

Artifact	Where it lives	Accessible?
Promoted docs	docs/features/<name>/	✅ Browsable in workspace
Feature history	docs/feature-history/<name>/	✅ Timeline view
Close summary	docs/features/<name>/close-summary.md	✅ Searchable
Audit trail	.dapei/runtime/audit/capability.log	✅ In-app audit view
Worktree (cleaned)	Removed from disk	⚠️ Recoverable from git if needed
Agent action log	.dapei/runtime/audit/	✅ Full detail

What is cleaned up

• The git worktree is removed from disk (you can still find the branch in your local git repo)
• The feature directory can be archived or removed
• Temporary runtime files

Why close matters

Closing a feature transforms ephemeral agent work into durable workspace knowledge. A closed feature is not "done and forgotten" — it's:

• Discoverable — feature-history makes it searchable
• Accountable — audit trail records who approved what and when
• Reusable — promoted docs feed into future features
• Navigable — cognitive index lets future agents learn from past work

How to browse feature history later

From the Workspace Browser, click Feature History or navigate to:

workspace
└── docs/
    └── feature-history/
        └── add-dark-mode-toggle/
            ├── index.md
            ├── business-design.md
            ├── technical-design.md
            ├── validation-report.md
            └── close-summary.md

Each history entry shows the decisions, test results, and key documents from that feature — a permanent record of what your AI agent built and how you guided it.

Next steps

• Feature Lifecycle — the full stage progression
• Stage Gates — gates that control advancement
• Audit Trail — read the audit log
• CDR Overview — how features feed into workspace knowledge