v0.1.0 — Pickle Rick for ForgeCode

Pickle Rick for ForgeCode v0.1.0

First release of the ForgeCode-native port. Not a wrapper around Claude Code — a ground-up rebuild using ForgeCode primitives.

What's Included

Orchestration (bin/)

• tmux-runner.js — Context-clearing iteration loop with phase-aware agent selection, parallel morty workers in git worktrees, circuit breaker, rate-limit detection
• microverse-runner.js — Metric-driven convergence loop with gap analysis, rollback, stall detection, LLM judge scoring, szechuan sauce + anatomy park sub-modes
• spawn-refinement-team.js — Parallel analyst spawner (3 workers/cycle, cross-reference, early exit)
• init-microverse.js — Session setup CLI
• setup.js — Session initialization

Core Libraries (lib/)

• state-manager.js — File-locked atomic state with transactions, crash recovery, schema migration
• circuit-breaker.js — 3-state FSM (CLOSED→HALF_OPEN→OPEN), 5-signal progress detection
• token-parser.js — auto_dump JSON role filtering with zero false-positive token detection
• git-utils.js — SHA-validated rollback, worktree management, cherry-pick with conflict detection
• handoff.js — Structured handoff generation (base + microverse + anatomy park variants)
• prd-gate.js — PRD verification readiness gate (section scan + quality scan)
• anatomy-park.js — Three-phase subsystem review engine

ForgeCode Agents (14 total)

Per-agent model selection, hard-enforced tool restrictions, persona injection for non-judges.

Skills (4)

Progressive disclosure with resource directories: microverse, anatomy-park, pickle, prd-draft.

Key ForgeCode Advantages

• forge -p --agent <id> — per-invocation agent selection with enforced tool restrictions
• auto_dump = "json" — full role discrimination for token detection (zero false positives)
• --cid — multi-turn conversation resume for PRD interviews
• Per-agent model selection — Haiku for judges, Sonnet for workers, Opus for complex analysis

Stats

• 562 tests, 0 failures
• 52 commits
• 7 lib modules, 5 bin scripts, 14 agents, 4 skills
• Platform verified via 12 spikes (10 runtime, 2 deferred low-risk)