SkillForge vs waza
A methodology and toolkit for engineering AI skills instead of improvising them — plus a Context Skill Advisor that proactively recommends, improves or creates skills from what you're doing. — versus — Microsoft's Go CLI for agent-skill quality: scaffold eval suites from SKILL.md, run them through Copilot, grade, compare models, replay snapshots and check spec and token budgets.
Both aim to make skill authoring an engineering discipline; skillforge is a methodology plus advisor, waza an eval-first CLI with graders, snapshots and CI gates.
| SkillForge | waza | |
|---|---|---|
| Stars | 905 | 1.4k |
| Forks | 93 | 92 |
| Language | Python | Go |
| License | MIT | MIT |
| Last activity | 2 months ago | yesterday |
| Topics | skills | skills, evals |
| Curated connections | 3 | 4 |
SkillForge — the curator's take
The 'skills are engineering, not art' manifesto, executed: structured creation process with validation built in, and v5.2's Context Skill Advisor watches your session/project context and suggests the skill you should have — with proactivity levels (off→active) so it advises rather than nags. If you author skills regularly, the rigor pays. NOT a registry or installer (that's the package managers' job) and the advisor's judgment tracks its context quality — garbage project context, garbage suggestions.
waza — the curator's take
Use it if you publish skills and want them tested like code — trigger-positive/negative tasks scaffolded from SKILL.md, shared graders, MCP and CLI mocks for deterministic CI, `spec verify` to prove the evals cover what the skill promises. Real runs go through the bundled GitHub Copilot CLI executor, so plan for Copilot access (a mock executor covers CI). Overkill for a personal skill you just iterate on by feel.