bkit v1.5.8 — PM Agent Team, Skills 2.0 & Team Orchestration

Highlights

bkit v1.5.8 is the largest feature release since v1.5.0, introducing the PM Agent Team for product discovery, Skills 2.0 Classification, and a complete Team Orchestration system with 5 coordination patterns.

PM Agent Team — Product Discovery Before Code

5 new specialized PM agents enable structured product analysis before writing a single line of code:

Agent	Role
pm-lead	Orchestrates 4-phase discovery workflow
pm-discovery	Opportunity Solution Tree (Teresa Torres)
pm-strategy	Value Proposition (JTBD) + Lean Canvas
pm-research	Personas, competitor analysis, market sizing (TAM/SAM/SOM)
pm-prd	8-section PRD synthesis with beachhead segment & GTM strategy

Run /pdca pm <feature> to generate a comprehensive PRD before planning.

Skills 2.0 — Smarter Skill Activation

All 35 skills now carry a classification tag — Workflow (W), Capability (C), or Hybrid (H) — enabling the skill orchestrator to make better activation decisions. 6 new skills added:

• /plan-plus — Brainstorming-enhanced planning with intent discovery and alternatives exploration
• /simplify — Code quality review and complexity reduction
• /batch — Process multiple features through PDCA in parallel
• /loop — Run commands on recurring intervals (e.g., /loop 5m /pdca status)
• /output-style-setup — Install output style files
• /pm-discovery — PM Agent Team product discovery workflow

Team Orchestration — Coordinated Multi-Agent Workflows

9 new modules in lib/team/ with 5 coordination patterns for Dynamic and Enterprise projects:

Pattern	Use Case
Leader	Standard feature development (1 lead + N workers)
Council	Architecture decisions (equal peer review)
Swarm	High-parallelism tasks (dynamic worker pool)
Pipeline	Multi-phase sequential workflows
Watchdog	Continuous monitoring with automated responses

Infrastructure

• Path Registry (lib/core/paths.js) — Centralized state file management with auto-migration from legacy paths
• Language Patterns (lib/intent/language-patterns.js) — Structured 8-language keyword patterns for intent detection
• 7th @import Module — executive-summary-rules.md for mandatory summary output after PDCA document work

---

What Changes for Users

Before v1.5.8

• Planning started immediately with /pdca plan — no structured product analysis
• Skills were flat — no classification, no activation intelligence
• Team mode was basic — limited coordination patterns
• 16 agents, 29 skills, 18 commands

After v1.5.8

• Product discovery first: /pdca pm <feature> generates a comprehensive PRD with personas, market analysis, and GTM strategy — then /pdca plan auto-references it
• Smarter skill suggestions: Skills are classified (W/C/H) so the system recommends the right skill for the right context
• Parallel team workflows: 5 orchestration patterns let multiple agents coordinate on complex features
• Enhanced planning: /plan-plus adds brainstorming phases (intent discovery, alternatives, YAGNI review) before document generation
• Automation tools: /loop for recurring monitoring, /batch for parallel multi-feature processing, /simplify for code quality passes
• 21 agents, 35 skills, 24 commands — significantly expanded capabilities

---

Component Counts

Category	v1.5.7	v1.5.8	Change
Agents	16	21	+5
Skills	29	35	+6
Commands	18	24	+6
Test Suites	24	78	+54
Test Cases	~218	972	+754
Lib Modules	~12	~25	+13
@import Modules	6	7	+1

Compatibility

• Gemini CLI: v0.29.0 ~ v0.33.x (tested through v0.33.0-preview.4)
• Claude Code: v2.1.34 ~ v2.1.71 (recommended: v2.1.71+)

Quality

• 972 / 972 tests passing (100%) across 78 test suites
• 11-perspective test coverage: unit, integration, E2E, scenario, philosophy, security, edge cases, boundary, recovery, compatibility, template
• 3 PDCA cycles completed at 100% match rate (implementation, testing, documentation)

---

Full Changelog: v1.5.7...v1.5.8