Skip to content

Sermilion/skill-bill

Repository files navigation

Skill Bill

A governed system for portable AI-agent behavior: stable skill entry points, shared orchestration, validator-backed contracts, cross-agent installers, scaffolding, workflow state, and local-first telemetry that keep one source of truth from drifting as the repo grows.

Skill Bill is a governance product, not a prompt dump. The product is the framework:

  • stable user-facing commands such as /bill-code-review and /bill-quality-check
  • manifest-driven platform packs under platform-packs/
  • shell/content contracts under orchestration/
  • runtime surfaces in the skill-bill CLI and skill-bill-mcp server
  • validator-backed authoring and scaffolding flows

Why it exists

Most prompt or skill repos degrade over time:

  • names drift
  • overlapping skills appear
  • stack-specific behavior leaks into generic prompts
  • different agents get different copies

Skill Bill treats skills more like software:

  • stable base capabilities
  • platform-specific overrides behind routers
  • shared contracts instead of prompt folklore
  • loud-fail validation instead of silent fallback
  • one repo synced across multiple coding agents

Core experience

Most daily use comes through a small set of stable commands:

  • /bill-code-review routes to the matching platform review stack
  • /bill-quality-check routes to the matching stack-specific checker
  • /bill-feature-implement orchestrates spec-to-PR work
  • /bill-feature-verify verifies a PR against a spec or design doc
  • /bill-pr-description generates PR text and QA steps

Skill Bill also ships:

  • authoring and maintenance skills such as /bill-create-skill, /bill-boundary-decisions, and /bill-boundary-history
  • a local CLI for telemetry, learnings, workflow resume, validation, scaffolding, and install primitives
  • an MCP server that exposes the same review, workflow, telemetry, and scaffolding primitives as agent tools

Quick install

git clone https://github.com/Sermilion/skill-bill.git ~/Development/skill-bill
cd ~/Development/skill-bill
./install.sh

The installer:

  • detects supported agents
  • installs selected skills as symlinks back to this repo
  • installs selected platform packs
  • registers the local Skill Bill MCP server for agents that support MCP config

Supported install targets today:

  • GitHub Copilot
  • Claude Code
  • GLM
  • OpenAI Codex
  • OpenCode

Start here

  • Getting Started: primary onboarding guide, install flow, skill surfaces, full skill-bill CLI coverage, and full MCP tool coverage
  • Getting Started for Teams: rollout guidance, customization strategy, trust-vs-verify guidance, and adoption patterns
  • Review Telemetry: telemetry contract, learnings, local DB usage, and remote proxy stats
  • Roadmap: product direction, priorities, and strategic framing

Reference skill catalog

Canonical Skills (13 skills)

Skill Purpose
/bill-boundary-decisions Record architectural and implementation decisions in agent/decisions.md
/bill-boundary-history Record reusable feature history in agent/history.md
/bill-code-review Stable code-review entry point that routes to the matching platform pack
/bill-create-skill Scaffold and bootstrap governed skills and platform packs
/bill-feature-guard Add feature-flag rollout safety to an implementation
/bill-feature-guard-cleanup Remove feature flags and legacy code after rollout
/bill-feature-implement End-to-end feature workflow from spec through review and validation
/bill-feature-verify Verify a PR against a task spec or design doc
/bill-grill-plan Stress-test a plan or design by walking the decision tree
/bill-pr-description Generate a PR title, description, and QA steps
/bill-quality-check Stable quality-check entry point that routes to the matching checker
/bill-skill-remove Remove an existing skill or platform skill set and clean up installs
/bill-unit-test-value-check Review unit tests for low-value or tautological coverage

Architecture snapshot

The main governed layers are:

  • skills/: canonical user-facing skills and pre-shell platform overrides
  • platform-packs/: manifest-driven platform review and quality-check depth
  • orchestration/: routing, delegation, workflow, telemetry, and shell-content contracts
  • skill_bill/: Python CLI, MCP server, workflow state, telemetry, scaffolding, and install primitives
  • scripts/: repo validators and migration helpers

Governed pack skills use a thin SKILL.md wrapper plus sibling content.md and shell-ceremony.md. Workflow shells such as bill-feature-implement and bill-feature-verify follow the same split: shell-owned orchestration and telemetry in SKILL.md, authored execution guidance in content.md.

Reference packs

The shipped reference inventory is intentionally narrow and deep:

  • kotlin: baseline Kotlin review and quality-check behavior
  • kmp: Kotlin baseline plus Android/KMP depth and governed add-ons

Routing, validation, and installation are manifest-driven, so the system is designed to accept any conforming pack rather than a hardcoded shortlist.

Validation

Core validation commands:

.venv/bin/python3 -m unittest discover -s tests
npx --yes agnix --strict .
.venv/bin/python3 scripts/validate_agent_configs.py

License

MIT

About

AI-agent skill governance system with manifest-driven routing, validator-backed contracts, scaffolding, installers, and local-first telemetry.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors