fix: /office-hours and /ceo-review accidentally contradict Boil the Lake (v0.16.1.1)

[simonemacario]

Summary

Users conflate "narrow scope" (product) with "cut quality" (implementation). gstack's own skills reinforce this confusion: /office-hours says "narrow beats wide" without distinguishing axes, and /plan-ceo-review SCOPE REDUCTION says "cut everything" without protecting implementation quality.

The fix: two axes, one mantra.

• Product scope (WHAT you build) — narrow early. Narrowest wedge, focus as subtraction.
• Implementation scope (HOW COMPLETELY you build it) — always maximal. Boil the Lake.

Narrow the product. Boil the implementation.

Changes (rebased onto v0.16.1.0)

1. ETHOS.md — New "Two Axes of Scope" section in the builder philosophy (renumber 2→3, 3→4)
2. scripts/resolvers/preamble.ts — Two-axis distinction added to generateCompletenessSection(). Propagates to ALL skills via {{PREAMBLE}}
3. office-hours/SKILL.md.tmpl — Q4 Narrowest Wedge: "this is PRODUCT scope, not implementation quality" + bonus distinction after Q4
4. plan-ceo-review/SKILL.md.tmpl — 5 surgical clarifications: SCOPE REDUCTION, Focus as subtraction, Subtraction default, scope application guidance, ruthless cut
5. All SKILL.md files regenerated + golden baselines updated

Pre-Landing Review

No issues found. Content-only change (Markdown text + one TypeScript string literal).

Test plan

• bun run gen:skill-docs regenerated cleanly for all 3 hosts (claude, codex, factory)
• Two-axis text propagates to all generated SKILL.md files via preamble
• Golden baselines updated (claude, codex, factory ship SKILLs)
• Rebased cleanly onto v0.16.1.0 (main as of 2026-04-08)
• CHANGELOG version sequence verified contiguous

🤖 Generated with Claude Code

[simonemacario]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

Verified: template-to-output pipeline (generateCompletenessSection(ctx) signature change + call site), commit bisection (4 logical commits), .tmpl editing workflow, and generated SKILL.md regeneration. All 557 tests pass per PR description.

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}

[simonemacario]

Rebased onto v0.13.0.0 — adapted to the new scripts/resolvers/ architecture (the original change targeted the monolithic gen-skill-docs.ts which has since been split).

The two-axis text now lives in scripts/resolvers/preamble.ts → generateCompletenessSection(), so it propagates to all skills via {{PREAMBLE}} as before. All SKILL.md files regenerated cleanly.

Updated the PR description to reflect the current state.

[simonemacario]

Rebased onto v0.14.1.0. One conflict resolution worth noting: the original commit 2 modified scripts/gen-skill-docs.ts directly, but upstream split that into scripts/resolvers/ since the last rebase. I dropped the stale commit 2 and kept commit 5 which already applies the two-axis text to scripts/resolvers/preamble.ts (the correct location). Net effect is the same — generateCompletenessSection() produces the two-axis guidance, all skills inherit it via {{PREAMBLE}}.

The gen-skill-docs.ts diff is large (-240/+216 lines) but it's entirely upstream's refactoring being accepted — no manual edits from this PR. The actual PR changes are confined to:

• ETHOS.md (+43 lines — new section)
• scripts/resolvers/preamble.ts (+4/-1 — the two-axis text)
• office-hours/SKILL.md.tmpl (+4/-1)
• plan-ceo-review/SKILL.md.tmpl (+10/-1)

[simonemacario]

Rebased onto current main (v0.16.1.0) — the original PR had merge conflicts from v0.14.1.0. Stripped out the structural changes that already landed (INVOKE_SKILL resolver, sensitive: true, factory simplification) and kept only the content changes: two-axis scope philosophy in ETHOS.md, preamble resolver, and surgical clarifications in office-hours + plan-ceo-review templates. All SKILL.md files regenerated cleanly for all 3 hosts.