ADR-0017: Claude skills scope includes domain scaffolding¶
- Status: Accepted
- Date: 2026-04-17
- Deciders: Lusoris, Claude (Anthropic)
- Tags: claude, agents, framework
Context¶
A minimal Claude-skill surface (build, test, lint) would not capture the fork's actual workflows. The user requested skills that scaffold new surfaces ("claude skills and hooks need more, like adding a new gpu backend or whatever"). Skills are the canonical entry point for repeatable multi-step tasks.
Decision¶
We will scope Claude skills to cover: build + test + domain scaffolding (add-gpu-backend, add-simd-path, add-feature-extractor, add-model) + profiling + bisect + upstream port/sync + release + format/lint.
Alternatives considered¶
| Option | Pros | Cons | Why not chosen |
|---|---|---|---|
| Build/test/lint only | Tiny skill set | Misses the scaffold workflows that motivated the fork | Rejected per user request |
| Full workflow coverage (chosen) | One canonical entry per repeated task | Larger skill inventory to maintain | User explicitly requested |
This decision was a default — the user specified the breadth.
Consequences¶
- Positive: every common fork workflow has a named skill; agents (human or AI) have predictable commands.
- Negative: skill set must be maintained and kept in sync with real flows.
- Neutral / follow-ups: ADR-0018 complements this with hooks.
References¶
- Source:
req(user: "claude skills and hooks need more, like adding a new gpu backend or whatever") - Related ADRs: ADR-0018