ship-loop（船环）

Ship Loop v5.0 — TARS Convergence

Orchestrate multi-segment feature work as a self-healing pipeline. Three nested loops ensure maximum autonomy: Loop 1 runs the standard code→preflight→ship→verify chain, Loop 2 auto-repairs failures via the coding agent, Loop 3 spawns experiment branches when repairs stall. A SQLite state backend provides crash recovery and cross-run analytics. A verdict router replaces hardcoded branching with a configurable decision table. A reflection loop audits historical effectiveness and auto-generates learnings.

Architecture: Three Loops + Event Queue + Verdict Router

┌───────────────────────────────────────────────────────────┐
│                  SHIP LOOP v5.0                           │
│                                                           │
│  LOOP 1: Ship Loop                                        │
│  code → preflight → ship → verify → emit(segment_shipped)│
│          │                                                │
│       on fail (verdict → action via VerdictRouter)        │
│          ▼                                                │
│  LOOP 2: Repair Loop                                      │
│  capture context → agent fix → re-preflight (max N)      │
│  ↳ emit events: repair_done | repair_failed               │
│  ↳ convergence detected → CONVERGED verdict → META        │
│  ↳ unknown error → record_decision_gap()                  │
│          │                                                │
│       exhausted                                           │
│          ▼                                                │
│  LOOP 3: Meta Loop                                        │
│  meta-analysis → N experiment branches → winner → merge   │
│  ↳ emit: meta_done                                        │
│                                                           │
│  🗄  SQLite (tars.db): runs, segments, events, learnings  │
│  📋  Event Queue: crash recovery via unprocessed events   │
│  🔀  Verdict Router: configurable verdict→action table    │
│  📚  Learnings Engine: scored lessons (score tracks use)  │
│  🪞  Reflect Loop: post-run analysis + recommendations    │
│  💰  Budget Tracker: token/cost tracking per run          │
└───────────────────────────────────────────────────────────┘

Security Notice

SHIPLOOP.yml is equivalent to running a script. The agent_command, all preflight commands (build, lint, test), and custom deploy scripts execute with your full user privileges. Ship Loop does not sandbox these commands. Never use on untrusted repos without reviewing the config. Treat SHIPLOOP.yml with the same caution as a Makefile or CI pipeline.

When to Use

• Building multiple features for a project in sequence
• Any work that follows: code → preflight → commit → deploy → verify → next
• When you need checkpointing so progress survives session restarts
• When you want self-healing: failures auto-repair before asking humans
• When you want cost visibility and learning from past runs

Prerequisites

• Python 3.10+ with pyyaml and pydantic installed
• A git repository with a remote
• A deployment pipeline triggered by push (Vercel, Netlify, etc.)
• A coding agent CLI configured via agent_command in SHIPLOOP.yml

Installation

pip install pyyaml pydantic

CLI Usage

# Core pipeline
shiploop run              # Start or resume the pipeline
shiploop run --dry-run    # Preview what would happen
shiploop status           # Show segment states (reads from DB)
shiploop reset <segment>  # Reset a segment to pending

# Learnings
shiploop learnings list
shiploop learnings search "dark mode theme toggle"

# Budget
shiploop budget           # Show cost summary

# v5.0 NEW
shiploop reflect          # Run meta-reflection on recent run history
shiploop reflect --depth 20  # Analyze last 20 runs
shiploop events           # View event history for latest run
shiploop events <run_id>  # View event history for specific run
shiploop history          # View past run history from DB

# Options
shiploop -c /path/to/SHIPLOOP.yml run
shiploop -v run           # Verbose logging
shiploop --version        # Show version (5.0.0)

Pipeline Definition (SHIPLOOP.yml)

project: "Project Name"
repo: /absolute/path/to/project
site: https://production-url.com
branch: pr               # direct-to-main | per-segment | pr
mode: solo

agent_command: "claude --print --permission-mode bypassPermissions"

preflight:
  build: "npm run build"
  lint: "npm run lint"
  test: "npm run test"

deploy:
  provider: vercel        # vercel | netlify | custom
  routes: [/, /api/health]
  marker: "data-version"
  health_endpoint: /api/health
  deploy_header: x-vercel-deployment-url
  timeout: 300

repair:
  max_attempts: 3

meta:
  enabled: true
  experiments: 3

budget:
  max_usd_per_segment: 10.0
  max_usd_per_run: 50.0
  max_tokens_per_segment: 500000
  halt_on_breach: true

# v5.0 NEW: Reflection config
reflection:
  enabled: true       # run reflect loop after pipeline
  auto_run: true      # automatically run, not just on CLI command
  history_depth: 10   # how many past runs to analyze

# v5.0 NEW: Custom verdict routing
router:
  agent_fail: retry      # override default (fail) with retry
  deploy_fail: fail      # override default (retry) with fail

segments:
  - name: "feature-name"
    status: pending
    prompt: |
      Your coding agent prompt here.
    depends_on: []

SQLite State Backend (v5.0)

State is now stored in .shiploop/tars.db (SQLite, WAL mode). SHIPLOOP.yml is config-only.

Tables

Event Types

Crash recovery: On startup, unprocessed events are replayed to restore pipeline state.

Verdict Router (v5.0)

The orchestrator no longer uses if/else chains. Every outcome maps to a Verdict, and a VerdictRouter maps verdicts to Action values.

Default Routing Table

Override via router: section in SHIPLOOP.yml (see above).

Meta-Reflection Loop (v5.0)

Runs automatically after pipeline completion (when reflection.auto_run: true) or manually via shiploop reflect.

What It Analyzes

1. Repeat failures — same error_signature across multiple segments/runs
2. Repair-heavy segments — segments that needed >1 repair loop (same error type)
3. Efficiency trends — cost/time per segment trending up or down
4. Stale learnings — learnings with score < 0.3 that haven't helped
5. Decision gaps — situations that triggered MISSING_DECISION_BRANCH

Auto-creates learnings from patterns

If an error signature appears 3+ times across runs, the reflect loop auto-generates a AUTO-<sig> learning flagging it for human review.

shiploop reflect --depth 20

═════════════════════════════════════════════════════
🪞  Ship Loop Reflection Report
   Generated: 2026-03-27T06:30:00Z
   Runs analyzed: 10
═════════════════════════════════════════════════════

📊 Efficiency
   Total cost:     $12.4200
   Segments run:   8
   Avg/segment:    $1.5525

🔁 Repeat Failures (2)
   abc123def456… × 3
   ...

💡 Recommendations
   ⚠️  Error signature abc123de… repeated 3× across segments: auth, api, db.
   📉 2 stale learning(s) (score < 0.3): L002, L004.
   ✅ No issues detected in recent history. Pipeline looks healthy!

═════════════════════════════════════════════════════

Playbook Evolution (v5.0)

When a repair fails with an error that doesn't match any existing learning, the system records a decision_gap:

learnings.record_decision_gap(
    segment="auth",
    context="Repair exhausted with unmatched error: ...",
    verdict="repair_exhausted_unknown_error",
    run_id="...",
)

Decision gaps surface in shiploop reflect output and the decision_gaps DB table. Operators use them to add new learnings or router overrides.

Convergence Detection (v5.0 Enhanced)

Same-segment: if two consecutive repair attempts produce the same error hash → CONVERGED verdict → router jumps to META (skipping remaining repair attempts).

Cross-segment: before starting a segment, the orchestrator checks if any already-shipped segment touched the same files (via touched_paths in DB). If overlap detected, a file_overlap_warning event is emitted.

Learnings Scoring (v5.0)

score (default 1.0)
  +0.1 when injected and segment succeeds first-try
  -0.2 when injected and segment fails the same way

Search results are sorted by combined keyword-relevance × score. Learnings with score < 0.3 are flagged as stale in reflection.

shiploop learnings list  # shows all learnings with scores

State Machine

States per segment:
  pending → coding → preflight → shipping → verifying → shipped
                  ↘ repairing (Loop 2) → preflight
                  ↘ experimenting (Loop 3) → preflight → shipping
                  ↘ failed

SHIPLOOP.yml checkpointed after every transition (for backward compat). SQLite is the primary state store.

Deploy Providers

Budget Tracking

Token usage and estimated costs tracked per agent invocation in SQLite (falls back to metrics.json).

shiploop budget

💰 Budget Summary: Portfolio
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Total cost:       $3.84
  Budget remaining: $46.16
  Total records:    12

  By segment:
    dark-mode: $0.42
    contact-form: $3.42
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Critical Rules

1. Never break the chain — after a segment ships, immediately start the next
2. Preflight is mandatory — no exceptions, no "ship now fix later"
3. Explicit staging only — never git add -A, only changed files from git diff
4. Prompts via file — never shell arguments (prevents injection)
5. SQLite is source of truth — SHIPLOOP.yml config-only; runtime state in tars.db
6. Agent command from config — always read from agent_command, never hardcode
7. Budget-aware — track costs, enforce limits, fail gracefully

Project Structure

skills/ship-loop/
├── SKILL.md                  # This file
├── pyproject.toml
├── shiploop/
│   ├── __init__.py           # __version__ = "5.0.0"
│   ├── cli.py                # CLI (run, status, reset, reflect, events, history, ...)
│   ├── config.py             # SHIPLOOP.yml parsing + validation (Pydantic v2)
│   ├── orchestrator.py       # Main state machine + event queue + verdict routing
│   ├── db.py                 # NEW: SQLite state backend (tars.db)
│   ├── router.py             # NEW: Verdict→Action router
│   ├── learnings.py          # Learnings engine (SQLite + scoring + decision gaps)
│   ├── budget.py             # Cost/token tracking (SQLite backend)
│   ├── git_ops.py            # git operations + get_touched_paths()
│   ├── agent.py              # Agent runner
│   ├── deploy.py             # Deploy verification
│   ├── preflight.py          # Build + lint + test runner
│   ├── reporting.py          # Status messages + reports
│   ├── ship_utils.py         # Ship and verify helper
│   └── loops/
│       ├── ship.py           # Loop 1: code → preflight → ship
│       ├── repair.py         # Loop 2: repair + decision gap detection
│       ├── meta.py           # Loop 3: meta-analysis + experiments
│       ├── reflect.py        # NEW: post-run reflection + recommendations
│       └── optimize.py       # Optimization loop
├── providers/
│   ├── vercel.py
│   ├── netlify.py
│   └── custom.py
└── tests/
    ├── test_config.py
    ├── test_orchestrator.py
    ├── test_git_ops.py
    ├── test_budget.py
    ├── test_learnings.py
    └── ...

Changelog

v5.0.0 (2026-03-27) — TARS Convergence

• SQLite state backend: tars.db replaces metrics.json + learnings.yml for runtime state
• Event queue: all phase transitions emit events; unprocessed events enable crash recovery
• Verdict router: configurable Verdict → Action table replaces if/else chains in orchestrator
• Meta-reflection loop: shiploop reflect analyzes run history, finds patterns, auto-generates learnings
• Playbook evolution: MISSING_DECISION_BRANCH detection → decision_gaps table
• Cross-segment convergence: touched_paths tracked per segment for overlap warnings
• Learnings scoring: score field (+0.1 on success, -0.2 on failure), sorted by score
• New CLI commands: reflect, events, history
• New config sections: reflection, router

v4.0.0

• Python CLI replaces bash scripts
• Pydantic v2 config validation
• Budget tracking with per-segment and per-run limits
• Error convergence detection (hash-based)
• Deploy provider plugins (Vercel, Netlify, Custom)