harness-proposelisted

# /harness-propose — trend-driven harness improvement proposal Survey the latest harness-engineering trends, judge fit against this repo, and visualize an improvement proposal as HTML. This is the **Guide generator** of the trust loop — it turns external signal + accumulated `/retro`·`/learn` findings into a concrete, reviewable backlog. It **proposes only**; it MUST pass a user approval gate before changing anything. > **Authority**: [`AGENTS.md`](../../../AGENTS.md) is binding. **MUST NOT** edit any `.claude/**`, `AGENTS.md`, > `CLAUDE.md`, or living report before explicit user approval. > **Reference loading (progressive disclosure):** > Read [principles](reference/principles.md) first, then use the [KB index](reference/kb/INDEX.md). > Load only topic files matching the current task; MUST NOT recursively load the entire KB. ## Project adaptation (read first) This repository may have **no** `.ignore/` workspace, `.claude/todo/`, `/harness` executor, or separate backlog file. In a lighter harness like this: - The proposal HTML is written to **`harness/proposals/<YYYYMMDD-HHmm>/index.html`** (git-ignored — derived view). This dated proposal **is** the detailed record of what was collected/judged. - Accepted items are reflected directly into the living **`harness/index.html`** "다음 단계" section (no separate backlog file — it was found redundant). There is no auto-executor — implementation is a normal, separately-approved task (often `skill-updater` for skills, plain edits f