v2.0.0 — Autonomous Agent Protocol

What's New

This release transforms the bootstrap from a scaffolding tool into an autonomous agent engine. Agents can now pick up work, execute, verify, and continue — without human prompting.

Based on lessons from two real products built with this guide (ai-plugin-manager, devlead-station), both with 400+ tests and complete implementations.

---

Autonomous Continuation Protocol

• Next Action table in the status dashboard — machine-readable, priority-ordered, per-hat
• Pick-up-work loop — 9-step protocol: read dashboard → find action → execute → verify → update → repeat
• Milestone state machine — not-started → in-progress → gates-passing → complete
• Quality gate commands — every milestone defines a concrete verification command; no vibes-based completion

Automated Validation Loop

• CEO hat spawns sub-agents wearing each persona to test real user scenarios end-to-end
• Fix → retest cycle repeats until 97% scenario success rate (configurable threshold)
• Each round re-tests ALL scenarios (catches regressions from fixes)
• Structured validation reports with pass/fail evidence per scenario

Comprehensive Test Strategy

Phase 8 (Testing & Quality Strategy) now prescribes a full taxonomy:

Functional: Unit, Integration, API/Contract, Component, System, E2E, Smoke, Sanity, Regression

Non-Functional: Performance (Load, Stress, Endurance/Soak), Security, Usability, Accessibility, Compatibility, Snapshot/Visual Regression

Each type specifies: framework, coverage target, CI vs manual, which milestone introduces it. Types can be skipped with explicit rationale.

CI/CD Bootstrapping

• Step 5 now scaffolds CI/CD config (.github/workflows/ci.yml)
• Minimum pipeline: lint → typecheck → test → build
• Multi-platform matrix guidance for desktop/native apps

Enhanced Hat Deliverables

• Business Analyst — research methodology: source verification, cross-referencing official docs
• UX Designer — HTML prototypes encouraged (interactive, self-contained)
• Developer (Build Plan) — milestone design: scope, dependencies, quality gates, done criteria
• QA — comprehensive test type checklist with per-milestone quality gates

New Anti-Patterns

• Milestones Without Quality Gates — no gate command = no state advancement
• Skipping Validation — all milestones complete ≠ product done; run the validation loop

Enhanced Templates

• Status Dashboard — now includes: Current Milestone, Next Action table, Milestones table with state + gate, Validation Status
• Root CLAUDE.md — includes Autonomous Continuation section with pick-up-work summary

---

Stats

• bootstrapping-guide.md: 1267 → 1469 lines (+16%)
• README.md: updated with autonomous continuation mention
• Zero bloat: every addition addresses a gap observed in real production use