ai-personal

Multi-agent software-engineering pipeline: a central Orchestrator dispatches Planner/Explorer/Coder/Reviewer/Arbiter/Executor spokes and a Scribe write-path service against one shared protocol.

9 agents

A two-axis orchestration lattice crossing model-cost team tiers (Opus orchestrator, Sonnet leads, Haiku workers) with T0-T3 tool escalation — Claude Code-native variant.

8 agents

A three-tier model-cost hierarchy — an Opus-tier core orchestrator plans and routes, Sonnet-tier research/coding/review team leads decompose and synthesize, Haiku-tier workers execute narrow tasks.

7 agents

Multi-session work loop: handoff.md carries state between sessions via resume-handoff/handoff-document, then ingest-handoff retires completed work into ADRs, docs/memory, and the handoff archive.

6 skills

Generate clean, atomic git commits from working changes with meaningful messages that explain why, not just what.

1 skill

Create a new Architecture Decision Record (ADR) in docs/adrs/

1 skill

Create or append a dated entry to a skill's CHANGELOG.md

1 skill

Run scripts/validate.sh, auto-fix mechanical structural issues, and report what needs manual authorship

1 skill

Plan multi-part work as a DAG of small slices with cost-appropriate models, then dispatch builders wave by wave

1 skill

Audit and refresh every README in the repo against a quality rubric, one reviewer subagent per workspace

1 skill

Validate, add, or update YAML frontmatter on documentation files

1 skill