takazudo/zudoesa-apply-voice

Apply esa Voice

Apply Takazudo's esa writing voice and vocabulary rules to incoming text.

Voice Character Summary

The esa voice is a casual memo/blog for colleagues. Like talking to a coworker — direct, relaxed, not overly polite. Fragment sentences are fine. The stance is "sharing" not "teaching".

Key traits:

• Casual directness — no roundabout preambles, get to the point
• Colleague-to-colleague tone — not customer-facing, not formal
• Self-reference: 「自分」or「Takazudo」. Never「筆者」(too formal for esa)
• Sentence endings: 〜なので、それ。/ 〜わけで / 〜という感じ / 〜かなと / 〜良さそう / 〜的な / 〜の模様
• Calm tone — casual but not emotionally dramatic or slangy
• Hedging via understatement: 〜の模様 / 〜良さそう / 〜かなと
• Avoid: 〜でございます / 〜させていただきます / 〜いただければ幸いです / 〜を説明する / 〜を解説する / overly polite language / stiff literary style

Contrast with CodeGrid voice

| Aspect | esa | CodeGrid | |--------|-----|----------| | Tone | Casual colleague memo | Polite but approachable tech article | | Formality | Low (断片的OK) | Medium (です/ます base) | | Self-reference | 自分 / Takazudo | 自分 / 筆者 / 私 | | Stance | Sharing with coworkers | Writing for readers | | Hedging style | Understatement (〜の模様, 〜良さそう) | Explicit softening (〜かと思います, 〜のではないでしょうか) | | Structure | Loose, memo-like | Considered, essay-like |

When to Use

• User provides text and wants it written/rewritten in Takazudo's esa voice
• User says "esa voice", "esa文体で", "esa風に書いて" etc.
• User wants to check or fix text to conform to the esa writing style

Workflow

Step 1: Read the rule files

Always read both rule files fresh at invocation time:

1. $HOME/repos/w/esa/doc/src/content/docs/overview/writing-style.md
2. $HOME/repos/w/esa/doc/src/content/docs/overview/vocabulary-rule.md

These files are the authoritative source of truth. Read them every time to pick up any updates.

Step 2: Identify the input text

The input text comes from one of:

• Text provided directly in the conversation (before or after the skill invocation)
• $ARGUMENTS passed with the command
• A file path the user points to

If no text is obvious, ask the user what text they want the voice applied to.

Step 3: Apply the rules

Transform or review the text using the voice character described above and the full details in the rule files. The rule files are the authority — the summary above is just for quick reference.

Step 4: Output the result

Output the transformed text. If the input was already close to the style, note what minor adjustments were made.

If reviewing existing text (rather than transforming), point out specific violations and suggest fixes.

Important Notes

• This skill transforms text style/voice only — it does not change the content or meaning
• The skill works on Japanese text. If English text is provided, translate to Japanese in the esa voice
• If $ARGUMENTS contains text, treat that as the input text to transform
• When in doubt about a style choice, refer back to the rule files — they are the authority
• The rules may be updated over time, which is why we re-read them every invocation