メニュー
Dev: check a target (spec / plan / working diff / code) for architectural drift from the rulebook docs/ARCHITECTURE.md — report qualitative deviations citing rule ids, and triage new ones into docs/Debt.md. Usage: /drift <target>
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | drift |
| description | Dev: check a target (spec / plan / working diff / code) for architectural drift from the rulebook docs/ARCHITECTURE.md — report qualitative deviations citing rule ids, and triage new ones into docs/Debt.md. Usage: /drift <target> |
| allowed-tools | Read Grep Glob Bash(git diff*) Bash(git status*) Edit |
Detect architectural drift: does $ARGUMENTS still fit how shadowling is built?
The target is a spec/plan path, the working diff (when omitted or changes), or a code
path / ..
Read the rulebook docs/ARCHITECTURE.md (rules carry ids R-<area>-<n>) and the known
deviations in docs/Debt.md. Judge the target only against the rules relevant to
it. The rulebook is the source of truth — do not restate or reinterpret it here.
Behavior:
R-id + file:line. No matching rule → not a finding —
don't free-associate. Bugs are /code-review; reuse/simplify cleanups are
/simplify.check.sh/tach already enforce (R-MOD, R-DEV-1/2/4) —
this gate is the qualitative layer.docs/Debt.md are known debt: list them as such, never re-raise
as new.docs/Debt.md ## Open in the file's format:
- [ ] R-<id> — <where>: <what deviates> → <target>. Record only what is confirmed.docs/ARCHITECTURE.md,
else drop. Never park a candidate.docs/Debt.md append
(deviation from an existing rule) and, on approval, a new rule in docs/ARCHITECTURE.md.
If the target matches the baseline and no gap surfaces, say so and skip triage.Push enriched vocab to Anki Desktop as flashcards and pull review progress back. Usage: /anki-sync
Add words/terms you're learning, enriched (translation, alternative translations, examples, synonyms, definition) into your native/learning languages. Comma-separated for several at once. Usage: /loot <word>[, <word2>, ...]
Review your buffered writing into per-category frequency docs (grammar / rephrasings / idioms / verbs). Usage: /debrief
Use whenever you read, query, or mutate shadowling's sqlite database (shadowling.db) — any SELECT/INSERT/UPDATE/DELETE, ad-hoc SQL, dropping/clearing/resetting rows, inspecting data, or checking a migration's result. All DB access goes through a skill entrypoint, a repository method (`models/*`), or `sql.py` — never raw sqlite3. Schema/data-layer changes are NOT made here — they go through an appended migration in appdb.py; this skill only documents that rule and verifies the outcome.
Explain an expression in the language you're learning that you can't read literally — get a verdict (memorize vs learnable rule) + how to read it, saved to the decode dataset. Usage: /aha <phrase> [+ your hunch]
Remove words/terms from your vocab list. Comma-separated for several at once. Usage: /drop <word>[, <word2>, ...]