Augments normal Claude Code conversations with language learning coaching for intermediate+ learners. Activate with `/learn-along <target-lang>` (e.g., `/learn-along en`). A subagent analyzes each exchange and provides natural expressions, vocabulary highlights, and corrections. Deactivate with `/learn-along off`. Triggers on "learn-along", "language coach", or requests to practice a language during coding sessions.
Augment everyday software engineering conversations with language coaching for intermediate+ learners.
/learn-along or /learn-along <target-language>When the user invokes /learn-along:
/learn-along en → target: English, native: auto-detect/learn-along from:ja to:en → native: Japanese, target: English/learn-along (no args) → target: English, native: auto-detect/learn-along off → deactivatelearn-along subagent.When native language is not explicitly specified:
/learn-along from:<native> to:<target> to adjust."When the user runs /learn-along off or says "stop language coaching":
/learn-along off to stop."IMPORTANT: Always complete the user's technical request fully before adding coaching.
Process the user's request and provide your full technical response as usual.
Use the Task tool with subagent_type: "learn-along" and model: "haiku". The subagent will classify the input and produce appropriate coaching output.
Provide in the prompt:
Example invocation:
Task tool call:
subagent_type: "learn-along"
model: "haiku"
prompt: |
Native: Japanese, Target: English
Domain: debugging test failures
User message:
このPRのテストが落ちてるから原因を調べて
Assistant response summary:
Investigated test failures, found mock setup issue in auth_test.rb, fixed by updating the expected arguments.
Append the subagent's output at the end of your response after a --- separator.
---
**Learn Along** (target: English)
[Subagent output here - varies by category]