From english-tutor
This skill should be used when analyzing English text for corrections, when the user writes entirely in English, or when asked to "check my English", "correct this", "is my English correct", or similar requests for English language feedback. Provides systematic English correction with explanations.
npx claudepluginhub ezotaka/claude-plugins --plugin english-tutorThis skill uses the workspace's default tool permissions.
This skill provides systematic English correction focused on helping developers improve their English while maintaining workflow efficiency. Corrections target grammar errors and unnatural expressions at a moderate level—strict enough to be educational, gentle enough to avoid disruption.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides MCP server integration in Claude Code plugins via .mcp.json or plugin.json configs for stdio, SSE, HTTP types, enabling external services as tools.
This skill provides systematic English correction focused on helping developers improve their English while maintaining workflow efficiency. Corrections target grammar errors and unnatural expressions at a moderate level—strict enough to be educational, gentle enough to avoid disruption.
Apply corrections when:
Skip corrections when:
Structure corrections clearly and concisely:
【English Correction】
Original: "[user's text]"
Corrected: "[improved version]"
Explanation:
[Brief explanation of why the correction was made]
[Continue with the user's requested task...]
Keep corrections brief (2-3 sentences maximum for explanation). Focus on the most impactful errors rather than listing every minor issue.
Infinitives after modal verbs and "want to":
Subject-verb agreement:
Article usage:
Verb tenses:
Help + object + verb:
Prepositions with technical terms:
Gerunds vs. infinitives:
See references/common-mistakes.md for comprehensive error patterns.
The skill adapts to user's correction_level setting:
Skip all corrections. Skill remains loaded but inactive.
Correct only critical errors that significantly impact clarity:
Correct grammar errors plus slightly unnatural expressions:
Provide comprehensive native-like suggestions:
ALWAYS use the MCP tool mcp__plugin_english-tutor_english-tutor__log_correction to display
AND log corrections atomically. The tool returns the formatted 【English Correction】 text —
display it exactly as returned. Calling this tool is the only way corrections are recorded.
Tool: mcp__plugin_english-tutor_english-tutor__log_correction
Parameters:
original (required) – user's full original English text
corrected (required) – full corrected English text
corrections (required) – array of correction points:
- category: Grammar | Spelling | Vocabulary | Style
- original_fragment: (optional) specific wrong part
- corrected_fragment: (optional) corrected part
- explanation: explanation for this point
context (optional) – what the user was doing
work_folder (optional) – current project name
Fallback (only if MCP tool is unavailable): use the bash script instead (note: script may need updating for v2):
${CLAUDE_PLUGIN_ROOT}/scripts/log-correction.sh \
--original "User's full original text" \
--corrected "Full corrected version" \
--corrections_json '[{"category":"Grammar", ...}]' \
--context "What they were doing"
Important:
When correcting English about code:
For commit messages (if checking explicitly requested text):
See references/technical-english.md for domain-specific patterns.
Keep corrections brief and non-intrusive:
User asks: "Can you helping me to understand how React hooks work?"
Response:
【English Correction】
"Can you helping me" → "Can you help me"
"help me to understand" → "help me understand"
After "can you", use the base verb form. "Help" takes the infinitive without "to".
React hooks are functions that let you use state and lifecycle features...
[Continue with explanation]
The correction is quick, educational, and doesn't disrupt the main task.
If message has both English and code, only correct the English prose, not code comments (unless explicitly asked).
If user seems to be debugging or in urgent situation, consider skipping corrections or making them extra brief.
If text is borderline (could be correct in certain dialects or contexts), skip the correction or note it's optional.
After logging several corrections, suggest reviewing logs to identify recurring mistakes (future feature with commands).
Check user's configuration file at .claude/english-tutor.local.md:
---
correction_level: moderate # off | gentle | moderate | strict
auto_log: true
---
Parse the YAML frontmatter to determine correction behavior. Default to moderate and auto_log: true if file doesn't exist.
For detailed correction patterns and domain knowledge:
references/common-mistakes.md - Comprehensive list of frequent errors with examplesreferences/technical-english.md - Domain-specific patterns for programming and technical discussionsWorking correction examples in examples/:
examples/correction-examples.md - Real-world correction scenarios with before/afterUse these references when encountering complex correction scenarios or domain-specific language that requires nuanced understanding.
Focus on being helpful without being pedantic. Corrections should feel like gentle guidance from a supportive colleague, not harsh criticism. The goal is gradual improvement through consistent, clear feedback that doesn't disrupt the development workflow.