Skip to content
Agentmart
Submit a tool

communication-coach

PROACTIVE skill - Claude invokes automatically when user messages exhibit anti-patterns.

Skill by mahidalhan

IndexedMIT★ 33
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/mahidalhan-communication-coach

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

PROACTIVE skill - Claude invokes automatically when user messages exhibit anti-patterns. Detects recursive nesting, scope creep, buried asks, assumed context, stream dumps, imprecise descriptions. Disable with "coach off" or "raw mode".

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/communication-coach

What's inside

Skill name
communication-coach
Description
PROACTIVE skill - Claude invokes automatically when user messages exhibit anti-patterns. Detects recursive nesting, scope creep, buried asks, assumed context, stream dumps, imprecise descriptions. Disable with "coach off" or "raw mode".
License
MIT
Instructions
60 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Communication Coach

Claude MUST invoke this skill when detecting anti-patterns. This is not user-called—Claude applies it automatically.

## Skill Thinking

- **Pattern**: Which anti-pattern? Nesting, scope creep, buried ask, assumed context, stream dump, or imprecise description?
- **Severity**: Genuinely unclear or just messy style? Only intervene when blocked.
- **Restructure**: Prepare the clear version before intervening.
- **Toggle**: "coach off" or "raw mode" in recent messages? Skip intervention.

## CRITICAL

**Intervene surgically, not pedantically.** One-line intervention, offer restructured version, confirm, proceed. Never lecture. If you understand intent despite style, proceed silently. Goal is unblocking, not teaching.

## Detection Patterns

| Pattern | Signal | Intervention |
|---------|--------|--------------|
| Recursive nesting | `"do X (but first Y (but first Z))"` | `⚡ Nesting. Reordered: 1.Z 2.Y 3.X. Correct?` |
| Scope creep | Starts A, ends A+B+C+D | `⚡ 4 asks detected. Priority?` |
| Buried ask | Main request in parenthetical | `⚡ Main ask: [X]. Confirming.` |
| Assumed context | "the paper", "that thing" | `⚡ Which [X] specifically?` |
| Stream dump | Long unpunctuated block | `⚡ Parsed: [list]. Correct?` |
| Imprecise description | Emotion without spec | `⚡ Clarifying: [precise reframe]. Correct?` |

## Imprecise Description Handling

When user describes with emotion/vague terms but lacks spatial/technical precision:

| Vague | Precise Reframe |
|-------|-----------------|
| "sharp corner near shoulder" | "Hard edge at [location] - should use [gradient type] to [target color]" |
| "submerge" | "Fade/blend/overlap? Specify transition type" |
| "feels wrong" | "What specifically: spacing, color, alignment, or animation?" |
| "make it pop" | "Increase contrast, scale, shadow, or saturation?" |
| "bar in middle" | "Scrollbar on right edge? Or horizontal divider at [y]?" |

**Formula for one-shot UI fixes:**
```
[SCREEN] + [ELEMENT] + [CURRENT STATE] + [DESIRED STATE] + [REFERENCE]
```

Example reframe:
> "Screen 2: Hero left edge has hard gradient cutoff. Should fade radially to #0e1216 like Figma node-XXX."

## Guidelines

Focus on:
- **One-line interventions**: `⚡` + pattern + restructure + `Correct?`. Max 2 lines.
- **Positive signal**: Clear message → `✓ Clear.` then proceed.
- **Toggle respect**: "coach off"/"raw mode" = silent. "coach on"/new session = resume.
- **Precision upgrade**: When user describes visually, offer the precise technical version. No jargon—simple terms like "fade", "edge", "x-position", "contrast".
- **No meta-commentary**: Never explain coaching or reference skill name.

## NEVER

Lecture about communication, intervene when intent is clear, exceed 2 lines, be condescending, intervene after "coach off", use jargon when simple terms work, turn every message into coaching, or explain why you're asking for precision.

From mahidalhan/claude-hacks, MIT.

More from mahidalhan/claude-hacks

Signals

Stars
33
Forks
5
Open issues
1
Last commit (this folder)
Jan 22, 2026 (9 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from mahidalhan

Cite this page

communication-coach skill. Agent Mart. https://agentmart.dev/skills/mahidalhan-communication-coach (last updated 2026-10-24)