Vocabulary Anti-Forgetting Skill

Use spaced repetition (based on the Ebbinghaus forgetting curve) to help the user memorize English phrases from the vocabulary bank.

Trigger

Activate this skill when the user says any of: 复习, 单词复习, review vocabulary, review words.

Files

| File | Purpose | |---|---| | vocabulary-anti-forgetting/asset/vocabulary_bank.md | Master list of 2087 English phrases with Chinese translations | | <workspace>/memory/review_log.md | Tracks every word's review history, level, and next review date (stored in the workspace memory folder, outside skill folder to survive upgrades) | | vocabulary-anti-forgetting/review.py | CLI script for non-interactive spaced-repetition review sessions |

Spaced Repetition Algorithm

Each word has a level that determines when it should be reviewed next:

| Level | Interval | Status | |---|---|---| | 0 | New word, never reviewed | new | | 1 | Review again in 1 day | learning | | 2 | Review again in 2 days | learning | | 3 | Review again in 4 days | reviewing | | 4 | Review again in 7 days | reviewing | | 5 | Review again in 15 days | reviewing | | 6 | Review again in 30 days | mastered | | 7 | Review again in 60 days | mastered |

Each time a word is shown (studied), it counts as a correct review: level + 1, schedule next review per the interval table.

Daily Review Session Flow (10 words)

When triggered, run the review script directly:

python vocabulary-anti-forgetting/review.py

This handles everything in one step: reads state, selects 10 words (due reviews first, then new words), displays them all at once, updates review_log.md with incremented levels and next review dates, and prints a session summary.

For custom word count: python vocabulary-anti-forgetting/review.py --count 5

Show the script output to the user as-is.

review_log.md Format

The file uses a markdown table with these columns:

| id | vocabulary | level | review_count | last_reviewed | next_review_date |

• id: matches the # column in vocabulary_bank.md
• vocabulary: the English word/phrase
• level: current spaced repetition level (0-7)
• review_count: total number of times reviewed
• last_reviewed: YYYY-MM-DD of last review
• next_review_date: YYYY-MM-DD when this word is next due

The file header also stores:

• total_sessions: running count of completed sessions
• last_session_date: date of the most recent session

Important Rules

• NEVER skip the update to review_log.md — this is the memory of the system.
• The review log lives at <workspace>/memory/review_log.md (the skill folder's parent memory/ directory), NOT in the skill folder. The script resolves this path automatically.
• If the user only wants a partial session (e.g., 5 words), that's fine — update only those words.
• If the user asks for stats (e.g., "复习统计"), read review_log.md and report: total words reviewed, mastered count, learning count, upcoming due words.
• The user can also run python vocabulary-anti-forgetting/review.py for a non-interactive session, or python vocabulary-anti-forgetting/review.py --stats for stats.