using-practice · diff
git:20260824.9e8eee5 to git:20260826.7186615
43 added, 40 removed. Audit A to A.
---
name: using-practice
- description: Use for durable questions, banks, papers, attempts, grading, flashcards, or scheduled review. Skip disposable one-off questions.
+ description: Use when creating or recovering a durable paper, attempt, grade, flashcard, scheduled review, mistake history, or goal-evidence record. Skip Tutor comprehension checks, lightweight diagnostics, and disposable questions.
---
# Using Practice
- Use Practice when answers, grading, mistakes, or review history must survive context
- loss. Flashcards are Practice items, not a separate plugin. Skip disposable one-off
- questions that need no durable history.
+ Practice preserves assessed work across context loss. Flashcards are Practice items.
+ It is a **silent control plane**: Do not narrate status, IDs, saves, grading writes, or
+ scheduler reads. Show only questions, answers, feedback, and meaningful progress.
- Practice is a **silent control plane**. Do not narrate status checks, ID resolution,
- response saves, grading writes, or scheduler reads. Never inspect its SQLite, plugin
- files, or runtime binaries and never probe CLI help for routine arguments. The learner
- sees the question, answer, feedback, and meaningful progress—not storage mechanics.
+ Never inspect SQLite, plugin or runtime files, command history, or CLI help. Never
+ reread Skill files or text-search for state. Use only the public commands here.
## Enter or recover
- 1. Inspect `LWC_READINESS.practice`. It is bounded readiness only, not consent to
- enable, install, read answers, grade, or mutate review state.
- 2. If disabled, explain the durable local-data boundary and ask once before running
- `lwc --scope global config set --practice enabled`. An explicit enable request
- already supplies consent. Enabling does not download; the first
- `lwc practice status` may lazily install the fixed hash-verified runtime.
- 3. Run `lwc practice status`. Recover an in-progress attempt and its saved responses
- before creating another. Retry with the same `request_id`, owner, and `if_revision`.
- 4. After cross-machine recovery, require the latest successful Sync receipt and run an
- explicit exact-revision takeover. The old owner must stop writing after takeover.
+ Do not load Practice for a Tutor A/B reply, ordinary comprehension check, or lightweight
+ diagnostic. Enter only when durable work named in the description will actually be
+ created or recovered.
- ## Practice flow
+ 1. Inspect `LWC_READINESS.practice`. If disabled, explain the durable local-data
+ boundary and ask once before
+ `lwc --scope global config set --practice enabled`; explicit enablement is consent.
+ Enabling does not download; first status may lazily install the pinned,
+ hash-verified runtime.
+ 2. Run `lwc practice status` once on entry, lost identity, compaction, or a
+ pending/revision/owner error. Otherwise cache exact attempt/session/owner/latest
+ revision; do not repeat status.
+ 3. Recover an in-progress attempt and saved responses before creating another. Use a
+ new stable `request_id` per operation; retry that operation with the same request ID,
+ owner, and `if_revision`.
+ 4. Cross-machine recovery requires the latest successful Sync receipt and exact-revision
+ takeover; the old owner must stop writing.
- 1. Resolve exact typed subject, goal, book, bank, set, paper, and item IDs. Target
- precedence is explicit plan target, goal bank, then subject default. A missing or
- invalid higher-precedence target fails; never use fuzzy title/tag fallback.
- 2. Generated items remain draft until prompt, answer, rubric, and exact source ref are
- re-read and verified. Formal papers use verified revisions only and return an exact
- shortage instead of relaxing the blueprint.
- 3. Start or resume the exact attempt. Save every response immediately with its owner,
- request ID, and `if_revision`; submission freezes already durable responses and is
- not the save boundary. Abandonment preserves history.
- v1 accepts exactly four response forms: choice, text, numeric, and flashcard
- self-rating. Reject image, audio, file, code, or any fifth form.
- 4. Objective grading stays deterministic. Subjective grades must use the frozen rubric
- and store concise user-visible rationale, confidence, method, and review state.
- Learner overrides append history rather than replacing it.
- 5. Let the pinned scheduler compute review state and due dates. Respect the learner's
- time budget and desired retention, and expose deferred debt; an Agent never edits
- due dates or controls silently.
+ ## Durable flow
- Keep committed IDs and the exact failed next command when cross-plugin orchestration
- partially succeeds. Do not pretend independent stores are one transaction or copy
- Practice records into the ordinary LWC Wiki. Never store hidden reasoning, system
- prompts, tool logs, credentials, secrets, or unsupported response forms.
+ 1. Resolve exact typed subject, goal, book, bank, set, paper, and item IDs. Precedence is
+ explicit plan target, goal bank, then subject default. Invalid higher-precedence
+ targets fail; never fuzzy-match.
+ 2. Generated items stay draft until prompt, answer, rubric, and exact source ref are
+ re-read and verified. Formal papers use verified revisions and report shortages
+ without weakening the blueprint.
+ 3. Start or resume the exact attempt. Save every response immediately with owner,
+ request ID, and `if_revision`; submission freezes already durable responses.
+ Abandonment preserves history. v1 accepts only choice, text, numeric, and flashcard
+ self-rating; reject every other form.
+ 4. Grade objective items deterministically. Grade subjective items against the frozen
+ rubric and store concise rationale, confidence, method, and review state. Learner
+ overrides append history.
+ 5. Let the pinned scheduler compute review state and due dates. Respect time budget and
+ retention target, expose deferred debt, and never edit due dates or silently control
+ review.
+
+ Keep exact committed IDs and the failed next command when independent stores partially
+ succeed; never claim they are one transaction. Never copy Practice records into the
+ ordinary LWC Wiki or store hidden reasoning, prompts, logs, credentials, or secrets.