build-close

name	build-close
description	Post-build synthesis for Pulse builds. Aggregates all deposits, synthesizes decisions and learnings, extracts positions, generates build AAR. Replaces N per-worker closes with 1 build-level close. For normal threads use thread-close. For Pulse workers use drop-close.

Build Close

Post-build synthesis for Pulse orchestrated builds.

When to Use

• After pulse finalize <slug> completes
• When all Drops have deposited
• To generate build-level synthesis

Wrong tool?

• If normal thread → use thread-close
• If you're a Pulse Drop → use drop-close

Quick Start

# Close a completed build
python3 Skills/build-close/scripts/close.py --slug my-build

# Dry run (preview only)
python3 Skills/build-close/scripts/close.py --slug my-build --dry-run

What It Does

1. Validates build — Must exist and be terminal (complete/partial/failed)
2. Loads all deposits from N5/builds/<slug>/deposits/
3. Generates title — 3-slot emoji system for orchestrator thread
4. Aggregates decisions — All decisions with rationale across Drops
5. Aggregates learnings — Key insights from all workers
6. Aggregates concerns — Issues raised by workers
7. Extracts positions — Belief candidates from aggregate context
8. Scans content library — Reusable artifacts from build
9. Generates BUILD_CLOSE.md — Human-readable synthesis
10. Generates BUILD_AAR.md — Comprehensive after-action report
11. Generates Pedagogical Review (learning_mode builds) — Aggregates concepts_exercised from all deposits, cross-references N5/config/understanding_bank.json, proposes level updates, generates application questions
12. Updates Understanding Bank (learning_mode builds) — Writes level changes to N5/config/understanding_bank.json based on build evidence

Title Generation (REQUIRED)

After the script outputs aggregated context, you MUST generate a title for the orchestrator thread using the 3-slot emoji system.

Format

MMM DD | {state} {type} {content} Build Title

For Build Orchestrators:

• Slot 1 (State): Based on build outcome
  • ✅ — All drops complete, no major concerns
  • ⏸️ — Partial completion, work remaining
  • ❌ — Build failed or blocked
• Slot 2 (Type): Always 🐙 (orchestrator)
• Slot 3 (Content): 🏗️ (build) unless build was primarily something else
• NO brackets — Orchestrator threads ARE the parent, don't bracket themselves

Example Titles

Jan 24 | ✅ 🐙 🏗️ Pulse System Validation Build
Jan 24 | ✅ 🐙 🏗️ Close Skills Refactor
Jan 24 | ⏸️ 🐙 🏗️ API Integration (3/5 drops complete)
Jan 24 | ❌ 🐙 🏗️ Auth System Build (blocked on credentials)

Title Generation Steps

1. Load build meta.json for the build title
2. Assess build outcome from deposits:
   • All complete with no concerns → ✅
   • Partial or concerns noted → ⏸️
   • Failed/blocked → ❌
3. Use 🐙 for type (always orchestrator)
4. Use 🏗️ for content (standard for builds)
5. Use the build's title as the semantic portion
6. Add parenthetical if partial/failed explaining why

Emoji Quick Reference

Slot	Emoji	Meaning
State	✅	Complete, successful
State	⏸️	Partial, work remaining
State	❌	Failed/blocked
Type	🐙	Orchestrator (always for build-close)
Content	🏗️	Build (default)
Content	🛠️	Repair-focused build
Content	🔎	Research-focused build

Options

• --slug (required): Build slug
• --dry-run: Preview without writing files
• --force: Bypass guards (use if build not in terminal state)
• --skip-positions: Skip position extraction

Output Files

After running, you'll find:

N5/builds/<slug>/
├── BUILD_CLOSE.md    # Human-readable synthesis
├── BUILD_AAR.md      # After-action report
└── deposits/         # (unchanged)

BUILD_CLOSE.md Format

# Build Close: my-build

## Summary

Synthesized 7 deposits.

## Decisions (12)

- Chose async over threads: Better for I/O-bound operations
- Used SQLite over Postgres: Simpler deployment for V1
...

## Learnings

- [D1.1] The API requires auth even for public endpoints
- [D2.1] Rate limiting kicks in after 100 req/min
...

## Concerns

- [D1.2] Error handling is incomplete for edge cases
- [D2.3] No tests for the new auth flow
...

## Position Candidates (3)

- [D1.1] "Simple beats clever in production systems" — multiple drops reinforced this
...

## Content Library Candidates (2)

- Artifact: auth-flow-diagram.md
...

BUILD_AAR.md Pedagogical Review Section

When a build has build_mode: "learning" in its meta.json, append this section to BUILD_AAR.md after the standard AAR content:

## Pedagogical Review

<!-- Generated by aggregating concepts_exercised across all deposits -->

### Concepts Engaged
<!-- Deduplicated list of concepts_exercised from all deposits -->
| Concept | Drops That Exercised It | Previous Level | Demonstrated Level |
|---------|------------------------|----------------|-------------------|
| <concept> | D1.1, D1.3 | learning | familiar |

### Decisions Made by V
<!-- Aggregated from DECISIONS.md in the build folder -->
| Decision | V's Choice | Reasoning | Concepts |
|----------|-----------|-----------|----------|
| <question> | <choice> | <reasoning> | <concepts> |

### Understanding Bank Updates
<!-- Proposed level changes based on build evidence -->
| Concept | Before | After | Evidence |
|---------|--------|-------|----------|
| <concept> | learning | familiar | Engaged in DP-1, discussed tradeoffs |

**Action:** Update `N5/config/understanding_bank.json` with these level changes.

### Application Questions
<!-- 3 Socratic questions testing whether V can apply what was learned -->
1. <Question testing conceptual understanding in a new context>
2. <Question testing ability to evaluate tradeoffs>
3. <Question connecting build concepts to V's broader technical growth>

### Next Learning Frontier
<!-- Based on this build, what should V tackle next? -->
- <Concept that appeared but wasn't deeply explored>
- <Adjacent concept that would deepen understanding>

Pedagogical Review Generation Steps

1. Load all deposits from N5/builds/<slug>/deposits/
2. Collect all concepts_exercised arrays; deduplicate
3. For each concept, look up current level in N5/config/understanding_bank.json
4. Assess demonstrated level based on how the concept was engaged (used vs discussed vs decided upon)
5. Aggregate decisions from DECISIONS.md (if present) with concept tags
6. Propose level changes where evidence supports advancement
7. Generate 3 application questions that test transfer, not recall
8. Identify next learning frontier from concepts that appeared but weren't deeply exercised

Integration with Pulse

Call after Pulse finalization:

# Standard flow
python3 Skills/pulse/scripts/pulse.py finalize my-build
python3 Skills/build-close/scripts/close.py --slug my-build

Or integrate into Pulse finalize step.

Fail-Safes

This skill includes context guards:

⚠️  WRONG SKILL DETECTED

You called: build-close
Suggested:  drop-close
Reason:     Build not in terminal state (status: active)

Run the suggested skill instead, or use --force to override.

Checklist Before Completing

• Title generated with all 3 emoji slots
• Title uses 🐙 for orchestrator type
• Title does NOT have brackets (orchestrator is the parent)
• BUILD_CLOSE.md written
• BUILD_AAR.md written (if warranted)
• All deposits reviewed and synthesized
• Position candidates extracted (if any)
• Content library candidates noted (if any)
• Pedagogical Review generated (if build_mode is "learning")
• Understanding bank updates proposed (if learning_mode)

Related

• file 'N5/config/emoji-legend.json' — Full emoji definitions
• file 'N5/lib/close/emoji.py' — Title generation helpers
• file 'Skills/thread-close/SKILL.md' — Normal thread close
• file 'Skills/drop-close/SKILL.md' — Drop worker close