slack voice

Developers often struggle with AI assistants that produce wordy, robotic Slack messages that scream "written by a bot." This skill fixes the register and formatting of outgoing communication to match a professional developer's natural voice. It enforces specific rules for different stakeholders, from peer engineers to VPs, ensuring your messages are brief and actionable.

What it does

• Register matching identifies the recipient (peer, manager, DevOps) and adjusts tone from technical deep-dives to high-level checks.
• Strict formatting removes common AI tells, bans em dashes, and varies sentence lengths to mimic human typing patterns.
• Drafting workflow creates native Slack drafts rather than auto-sending, allowing for final human review.
• Adaptive learning analyzes how you rewrite drafts to update its internal style guide and avoid repeating rejected patterns.
• Bilingual enforcement ensures messages use exactly one language, automatically switching to English if technical terms or ticket keys are present.

How it works

1. Context sampling scans the current thread or DM to understand the existing conversation flow and participant roles.
2. Drafting generates a message based on one of eight specific register types, such as "Infra/DevOps ask" or "Peer day-to-day."
3. Refinement compares the draft against a checklist to strip out signatures, bot mentions, and forbidden "AI-tell" words.
4. Learning triggers when you edit a draft, performing a diff to extract style lessons for future messages.

Frameworks & tools

This skill requires a Slack integration with permissions to read message history and create drafts. It is designed for use within LLM environments that support long-term memory or file-based updates to track writing preferences.

Why this beats prompting it yourself

Standard prompts often result in "bullet walls" and hedging that feel out of place in a fast-paced engineering Slack. This skill encodes specific social engineering rules, like placing proof links last and strictly separating English from other languages to avoid awkward calques.

Use cases

• Drafting a concise request to DevOps for a secret rotation without friction.
• Explaining a complex auth mechanism to a peer using technical specifics.
• Providing a high-level status check to a VP without a process dump.
• Redirecting a query to the correct owner with a warm, one-line handoff.

Known limitations

It cannot join huddles or participate in active DMs unless explicitly asked. The skill does not support mixing languages within a single message.