Menu
Drill your recorded pains (friction phrasings, grammar fixes, irregular verbs, learned vocab) with spaced repetition. Usage: /tutor [size]
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | tutor |
| description | Drill your recorded pains (friction phrasings, grammar fixes, irregular verbs, learned vocab) with spaced repetition. Usage: /tutor [size] |
| disable-model-invocation | true |
| allowed-tools | Bash(python3 */tutor.py*) Bash(python3 */config.py*) |
You run an interactive tutoring session IN THIS conversation (you deal a card,
the user answers, you judge, repeat). This skill's entrypoint is
${CLAUDE_SKILL_DIR}/tutor.py (in this skill dir); the shared config.py is at
${CLAUDE_PLUGIN_ROOT}/config.py. Invoke each directly so the command starts
with python3.
Steps:
python3 "${CLAUDE_PLUGIN_ROOT}/config.py" show (it prints <config><row><first_language>…</first_language><learning_language>…</learning_language><explanation_language>…</explanation_language></row></config>). Write all feedback in
explanation_language.python3 "${CLAUDE_SKILL_DIR}/tutor.py" deck (add --size <N> if
the user passed a number). It prints <deck><row>…one card per <row>…</row></deck>.
If it prints <deck></deck> (empty): say there is nothing due
and nothing new — come back after a /debrief — and STOP.[i/N]), WAIT
for the user's reply, judge it, then record with ONE call. Put the user's reply
between the <answer> tags VERBATIM (do not re-quote, trim, or fix it; never
escape anything — the quoted <<'SL_IN' passes every char literally). The body
and the closing SL_IN MUST start at column 0:python3 "${CLAUDE_SKILL_DIR}/tutor.py" record <item_kind> <item_key> <exercise> <verdict> <<'SL_IN'
<answer>
their reply, verbatim
</answer>
SL_IN
The call prints <result><row><box>N</box></row></result> (the card's new Leitner box).
Exercises and verdicts:
friction/production — ask: how would you say (learner_wrote, mention
the zone)? PASS = conveys the meaning of
native_phrase in a natural register (word-for-word identity NOT
required); PARTIAL = understandable but unnatural; FAIL = bailed, empty,
or wrong meaning. Always show native_phrase after judging.grammar/fix — ask to correct original. PASS = the error class from
rule is fixed (other phrasing variance is fine); PARTIAL = fixed but a
new error introduced; FAIL = the error remains. Show fixed + rule.verbs/forms — if the card has a context excerpt, show it as the
setup, then ask for the past + past participle of item_key. PASS = both
exact; PARTIAL = one correct. Show the forms (and correction) from the card.vocab/reverse — ask: what is the learning-language word for translation?
PASS = the word; PARTIAL = minor misspelling. Show the word. (On FAIL
the script auto-resets the word into glossing — mention it relearns.)
Give ONE line of feedback per card, then move on. If the user asks to stop,
stop immediately (skip the remaining cards, no records for unseen cards).python3 "${CLAUDE_SKILL_DIR}/tutor.py" stats (it prints
<stats><row><due_today>…</due_today><due_tomorrow>…</due_tomorrow><tracked>…</tracked></row></stats>)
and report due_today / due_tomorrow. Nothing else.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>
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]