원클릭으로 Manus에서 모든 스킬 실행

adr-scan

스타0

포크0

업데이트2026년 7월 11일 09:17

PROVISIONAL / one-off. Read-only full-repository scan that discovers ADR-worthy architectural decisions across the whole repo and maps each to the BACKLOG frame-ID taxonomy (Tier 0–6 / G,T,R,A,B,C,D). Unlike the go-boilerplate original (which migrated a flat decisions.md into docs/adr/), this repo already has a formal docs/adr/ set + BACKLOG.md board — so the job here is DISCOVERY of decisions that exist de-facto (in AGENTS.md, config files, src/ structure, .github/, code comments) but are NOT yet tracked as a BACKLOG frame, classifying each as decision (frame-worthy) / exclusion (frame-worthy negative decision) / rule (stays in AGENTS.md) / inventory (living reference), and proposing which Tier / frame ID it belongs to (existing or new). Read-only: produces a candidate inventory only; writes no docs/adr files and does not edit BACKLOG.md. Delete or archive once the discovered gaps are folded into BACKLOG.md.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

Tomy-ch

Tomy-ch/nextjs-boilerplate

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

adr-scan (provisional)

Status: temporary / one-off. Built to sweep the repo for architectural decisions that are real but untracked — deciding whether each deserves a slot on docs/adr/BACKLOG.md. Read-only: it emits a candidate inventory only. It never writes docs/adr/**, never edits BACKLOG.md, never edits source. Remove or archive it once the discovered gaps have been reconciled into the BACKLOG.

Why this exists here (differs from the go-boilerplate original)

The go-boilerplate version migrated a flat docs/decisions.md into a formal docs/adr/ set. This repo is already past that: it has a formal ADR set (docs/adr/0001-* … Dev-000x … Toolchain-000x) and a live board (docs/adr/BACKLOG.md) that tracks every decision area across Tiers 0–6 with frame IDs (G / T / R / A / B / C / D) and a 選定済み / 実装済み status pair.

So the useful job here is gap discovery: find decisions that are being made de facto — encoded in AGENTS.md, config files, the src/ layout, .github/, or code comments — but are not represented by any BACKLOG frame (or are represented but mis-tiered / mis-classified). The output feeds the BACKLOG運用ルール flow: 新しい意思決定領域に気付いたら該当 Tier への追加 → 内容合意 → ADR 化.

Classification taxonomy (the core judgment)

Each candidate is exactly one of:

decision — a choice among alternatives with lasting consequences (X over Y). BACKLOG-frame-worthy.
exclusion — a deliberate "we intentionally do NOT do X" with rationale. Frame-worthy (negative decision; the BACKLOG "明示的に boilerplate では決めない (out of scope)" section is the home for some of these).
rule — a day-to-day constraint/consequence (e.g. "use the @/* alias", "commit prefixes"). Stays in AGENTS.md; may reference an ADR but is not itself a new decision area.
inventory — a catalog that drifts with code (dependency list, script list). Stays in a living reference doc / README, never an ADR.

A candidate is frame-worthy only if: it has (or implies) considered alternatives, it is cross-cutting or hard to reverse, and it is not merely restating an existing rule/inventory or an already-tracked frame.

Scan surfaces (fan out one read-only worker each)

Fan these out in parallel via the Agent tool (read-only); each returns candidates in the output shape below. Then the orchestrator dedups against the existing BACKLOG frames.

Existing ADR + board — docs/adr/*.md + docs/adr/BACKLOG.md. Build the baseline set of already-tracked frames (so discoveries can be diffed against it). Note any ADR whose Status or BACKLOG status pair looks inconsistent with reality.
AGENTS.md — the ## [TODO] sections (each is a pending area — confirm it maps to a BACKLOG frame), the "AI Modification Scope" / "Protected Documentation" / "Git Rules" / "Language Rules" sections. Separate genuine decisions from rules.
Config & tooling (latent decisions) — package.json (deps, scripts, packageManager), tsconfig.json, next.config.ts, biome.json, postcss.config.mjs, mise.toml, .makefiles/**, .github/**. A pinned tool, an enabled compiler flag, a CI job, a browserslist — each may be a decision made in config but never elevated to an ADR/frame.
src/ de-facto structure — the actual directory layout, "use client" placement, route conventions, styling approach (Tailwind usage), any state/data-fetch pattern. These are the A-tier / B-tier de-facto states; check each is reflected (as ⚠️ de-facto) in BACKLOG and not silently hardening into an unrecorded convention.
Latent in prose/comments — README*, code // TODO / // why comments, .github/copilot-instructions.md. Decisions or exclusions stated in passing but never tracked.

Output (per candidate)

- title:         short decision title (imperative, ADR-style)
  type:          decision | exclusion | rule | inventory
  frame_worthy:  yes | no
  source:        file:line (evidence)
  alternatives:  present | implied | none
  backlog_state: tracked <frameID> | untracked | mis-tiered <current→proposed> | mis-classified
  proposed_tier: Tier N + letter (e.g. "Tier 4 / B") — existing frame ID if one fits, else "new"
  note:          one line — why worthy / why not / what's inconsistent

Aggregate into:

(A) Untracked decisions/exclusions — frame-worthy but absent from BACKLOG, with a proposed Tier + frame ID. This is the primary deliverable.
(B) Tracked-but-drifted — frames whose BACKLOG status pair (選定済み / 実装済み) or Tier looks wrong vs the observed reality.
(C) Stays-a-rule list — belongs in AGENTS.md, not a new frame.
(D) Stays-inventory list — belongs in a living reference, never an ADR.

The orchestrator presents these to the user as candidates only. Folding any of them into BACKLOG.md (or filing an ADR) is a separate, user-approved step — this skill does not perform it (BACKLOG.md is AI-editable per CLAUDE.md, but the 運用ルール requires 内容合意 before a frame is added).

Constraints

✅ Read-only. Emit a candidate inventory only. Never write docs/adr/**, never edit BACKLOG.md or source, never file an ADR.
✅ Respect the Instruction Priority: AGENTS.md > ADRs > BACKLOG > agent configs. A "decision" that contradicts an Accepted ADR is a finding, not something to normalize.
✅ Provisional: delete or archive this skill once the discovered gaps are reconciled into BACKLOG.md.
❌ Do not invent decisions to fill perceived gaps — only surface what has real evidence (file:line).

이 저장소의 다른 Skills

같은 저장소

node-upgrade

Tomy-ch/nextjs-boilerplate

Upgrade the Node.js version used by this project. `mise.toml` `[tools] node` is the single source of truth (ADR 0003); this skill confirms the target version with the user via AskUserQuestion, reviews the release notes / breaking changes for that Node line, edits `mise.toml`, has the user run `make install-tools` (mise install), then rebuilds the lockfile and verifies with `pnpm install` + `pnpm lint` + `pnpm build`. Unlike go-boilerplate's go-upgrade there is no `make sync-versions` / Dockerfile / go.mod to propagate to (ADR 0004 no-docker), and no CI node-version to sync yet (BACKLOG B9 pending). A `@types/node` major realignment is intentionally NOT bundled here — it is a separate PR per Toolchain-0005 (major updates in isolation). Use this for a deliberate Node version move; for a routine bulk audit of all mise tools (node + pnpm + …) with supply-chain quarantine, use `tools-upgrade` instead.

2026-07-110

repo-ops

Tomy-ch/nextjs-boilerplate

Operational runbook for this repository's recurring, easy-to-trip-on gotchas around the mise-managed toolchain, the pnpm lockfile, the Makefile setup targets, and the review-artifact directory. Read-only knowledge skill — it tells you the exact command to run; it does not silently mutate state. This is deliberately a SPARSE STARTER for the Next.js boilerplate: it carries only the gotchas that genuinely exist today (mise / pnpm / make DRY_RUN / tmp-reviews), and grows as new operational traps are discovered (in contrast to the go-boilerplate original, whose items were mostly Docker / sqlc / DB-runner specific and do not apply here — ADR 0004 no-docker). Triggers: "make install-tools が mise not found で落ちる", "DRY_RUN=0 なのに dry-run になる", "pnpm install --frozen-lockfile が落ちる", "mise.toml を変えた後の反映", "tmp/reviews が git に乗る".

2026-07-110

full-apply

Tomy-ch/nextjs-boilerplate

The "apply" skill that applies = fixes the findings under tmp/reviews/ generated by full-verify (architecture.md / mod_*.md / _index.md), top-down by severity (Critical → High → Medium → Low). The counterpart to the read-only full-verify. Skips (defers) "suspicious" findings that require design judgment or whose behavior is unclear, leaving a reason. Tracks done/deferred in the tmp/reviews/working.md ledger and stamps a status comment + commit hash at the top of each mod_*.md. Conflicts are first-come-first-served; within one md, related findings are fixed together within the bounds of not breaking the public API. Verifies (pnpm fix / lint / build, plus tests when present) before committing. Use it when asked to "fix the review findings", "apply reviews", or "full apply".

2026-07-110

full-verify

Tomy-ch/nextjs-boilerplate

Verify a whole repository's architecture and the validity of all implementation code in the background, generating a set of Markdown findings (tmp/reviews/architecture.md / mod_*.md / _index.md). The skill itself detects and adapts to the language, structure, and presence of design documents. Use it when asked for a whole-repository structural verification / implementation-soundness review / overall review / full verify — NOT a diff review. Changes no code; read-only plus Markdown generation only.

2026-07-110

commit

Tomy-ch/nextjs-boilerplate

Analyze the current working-tree changes (staged + unstaged), group them into appropriately-scoped commits with the project's prefix convention (Feat / Fix / Refactor / Perf / Docs / Test / Build / CI / Chore / Style / Revert), and execute each commit in Japanese after user approval. Pre-flight also checks whether the current branch's PR is already merged and, if so, recommends cutting a fresh branch from the base before committing. Commits are made with `git commit --no-verify` to skip lefthook during the split; after all commits succeed, the command runs the lefthook-defined commands directly plus `make fix` as a final verification gate (lefthook itself is bypassed because it skips checks when nothing is staged). Respects CLAUDE.md's git rules (no direct commits to protected branches, no force-push, no auto-push after PR amend, Co-Authored-By footer, HEREDOC commit messages).

2026-07-110

submit-pr

Tomy-ch/nextjs-boilerplate

Push the current feature branch to `origin` and create or update its GitHub pull request. Detects whether a PR already exists for the current branch via `gh pr view` and automatically chooses between "create" and "update". The PR body is filled from `.github/pull_request_template.md` (sections `概要` / `変更内容` / `動作確認方法`) using the commit history and diff. Title and body are written in Japanese per `CLAUDE.md`. The skill confirms with the user before any push, with the exact wording required by `CLAUDE.md` for the update path.

2026-06-070

name

adr-scan

description

adr-scan (provisional)

Why this exists here (differs from the go-boilerplate original)

Classification taxonomy (the core judgment)

Each candidate is exactly one of:

decision — a choice among alternatives with lasting consequences (X over Y). BACKLOG-frame-worthy.
exclusion — a deliberate "we intentionally do NOT do X" with rationale. Frame-worthy (negative decision; the BACKLOG "明示的に boilerplate では決めない (out of scope)" section is the home for some of these).
rule — a day-to-day constraint/consequence (e.g. "use the @/* alias", "commit prefixes"). Stays in AGENTS.md; may reference an ADR but is not itself a new decision area.
inventory — a catalog that drifts with code (dependency list, script list). Stays in a living reference doc / README, never an ADR.

Scan surfaces (fan out one read-only worker each)

Fan these out in parallel via the Agent tool (read-only); each returns candidates in the output shape below. Then the orchestrator dedups against the existing BACKLOG frames.

Existing ADR + board — docs/adr/*.md + docs/adr/BACKLOG.md. Build the baseline set of already-tracked frames (so discoveries can be diffed against it). Note any ADR whose Status or BACKLOG status pair looks inconsistent with reality.
AGENTS.md — the ## [TODO] sections (each is a pending area — confirm it maps to a BACKLOG frame), the "AI Modification Scope" / "Protected Documentation" / "Git Rules" / "Language Rules" sections. Separate genuine decisions from rules.
Config & tooling (latent decisions) — package.json (deps, scripts, packageManager), tsconfig.json, next.config.ts, biome.json, postcss.config.mjs, mise.toml, .makefiles/**, .github/**. A pinned tool, an enabled compiler flag, a CI job, a browserslist — each may be a decision made in config but never elevated to an ADR/frame.
src/ de-facto structure — the actual directory layout, "use client" placement, route conventions, styling approach (Tailwind usage), any state/data-fetch pattern. These are the A-tier / B-tier de-facto states; check each is reflected (as ⚠️ de-facto) in BACKLOG and not silently hardening into an unrecorded convention.
Latent in prose/comments — README*, code // TODO / // why comments, .github/copilot-instructions.md. Decisions or exclusions stated in passing but never tracked.

Output (per candidate)

- title:         short decision title (imperative, ADR-style)
  type:          decision | exclusion | rule | inventory
  frame_worthy:  yes | no
  source:        file:line (evidence)
  alternatives:  present | implied | none
  backlog_state: tracked <frameID> | untracked | mis-tiered <current→proposed> | mis-classified
  proposed_tier: Tier N + letter (e.g. "Tier 4 / B") — existing frame ID if one fits, else "new"
  note:          one line — why worthy / why not / what's inconsistent

Aggregate into:

(A) Untracked decisions/exclusions — frame-worthy but absent from BACKLOG, with a proposed Tier + frame ID. This is the primary deliverable.
(B) Tracked-but-drifted — frames whose BACKLOG status pair (選定済み / 実装済み) or Tier looks wrong vs the observed reality.
(C) Stays-a-rule list — belongs in AGENTS.md, not a new frame.
(D) Stays-inventory list — belongs in a living reference, never an ADR.

Constraints

✅ Read-only. Emit a candidate inventory only. Never write docs/adr/**, never edit BACKLOG.md or source, never file an ADR.
✅ Respect the Instruction Priority: AGENTS.md > ADRs > BACKLOG > agent configs. A "decision" that contradicts an Accepted ADR is a finding, not something to normalize.
✅ Provisional: delete or archive this skill once the discovered gaps are reconciled into BACKLOG.md.
❌ Do not invent decisions to fill perceived gaps — only surface what has real evidence (file:line).