The router plus the enforcement layer. Skills are inert without the router (self-trigger measured at 0/3), so this plugin ships both: the trigger matrix and Iron Laws, and the hooks that re-inject the floor after compaction and classify each prompt against the three MANDATORY domains.
Full async multi-agent development workflow: Planning → Design → Dev → Security/Tech Review loop with file-based handoffs and release-branch Git model.
Worker skills and agents: the issue worker, the fresh-context reviewer panel, the test hunt and the Claude Code documentation lookup; the ameise controller drives the stages.
AI-powered QA automation and Git workflow toolkit. /aqa-inspect runs end-to-end QA (generate test cases from a Figma design or live-URL exploration → execute → results.csv + HTML report), /aqa-jira files Jira tickets from failed QA results behind a human gate, /create-pr creates pull requests, /merge-check does dry-run merge validation, /explain-concept builds a two-layer interactive HTML explainer for one AIPub topic.
Vibe Driven Development role skills: brainstormer, wayfinder, start-loop, planner, orchestrator, plan-reviewer, coder, code-reviewer, create-pr. The Planner ends in a spec and tickets the Orchestrator hosts a Coder to implement.
The shared half of the AI-native SDLC: skills that write intent.md, spec.md and plan.md; verifier, simplifier and researcher agents; user-invocable skills for each stage gate and an /adopt skill for existing repos; and guardrail hooks that protect tests during a fix and keep secrets out of the diff.
Portable agentic-coding extensions: slash commands, subagents, and skills for Claude Code, with /optimise auditing a project's Claude Code, Claude Design, or Codex setup.
Export, import, and sync a Multica workspace — settings, skills, agents, squads, projects, autopilots, labels, properties, and the MCP roster — between workspaces.
Universal folder-based team configuration system. Save, load, create, and reuse multi-agent team packs that work with any LLM runtime — with 3 built-in team packs for software dev, marketing, and devops.
Minimal Exolvra Genesis loop orchestrator — builder subagents iterate against blind, fresh-context critics until the work wins the comparison twice in a row.
Validate OpenSpec or SpecKit change packages (proposal.md, spec.md, tasks.md) against a repository's real Makefile targets, coverage floor, waivers and witnesses using the deterministic planlint CLI. Use when asked to validate, lint, check or repair specs or implementation plans, before writing code from a plan, or when a spec-gate CI job fails. Read-only by default; never writes waivers or witnesses.
Claude Code skills for planning, adversarial and pull-request review, mutation testing, task tracking, and observability. Installed as one plugin; each skill is invoked as /lunar:<name>.