fullcheck

name	fullcheck
description	Comprehensive writing review — spawns analyser agents for thorough quality and voice assessment. Use when the user asks for a "full check", "thorough review", "comprehensive edit", or wants detailed quality analysis of their writing.
argument-hint	[file/dir/text]
disable-model-invocation	true

/fullcheck — Comprehensive Writing Review

Thorough quality review using writing-analyser subagents. Use this for important documents that need a proper edit.

Determine Input

Figure out what to check:

• File path(s) passed as arguments → use those files
• Directory path → use the files in it (markdown/text files)
• Inline text → use it directly
• Nothing passed → look at recent conversation context for text being discussed. If genuinely unclear, ask the user what they'd like checked.

Spawn Analyser Agents

Launch writing-analyser subagents in parallel using the Task tool. Split the work sensibly:

For a single document:

• Agent 1: Focus on nonos.md compliance and general writing quality (slop, clarity, word choice, rhythm)
• Agent 2: Focus on Danny's voice, structure, and overall effectiveness (voice compliance, structural patterns, document flow)

For multiple documents:

• One agent per document (up to 3-4), each doing a full review

Give each agent clear instructions about what to focus on and pass the file path(s) or text.

Compile Results

Once the agents return, compile their reports into a single comprehensive review:

1. Deduplicate — if both agents flag the same issue, list it once
2. Prioritise — order by severity (critical → warnings → suggestions)
3. Summarise — add a brief overall assessment at the top

Output Format

## Overall Assessment
[2-3 sentences: what's the text doing well, what needs the most work]

## Critical Issues
[Must fix — slop, errors, broken clarity]

## Warnings
[Should fix — weak language, structural problems, voice misses]

## Suggestions
[Nice to have — polish, enhancement opportunities]

## CLI Tool Output
[If agents ran write-good/proselint, include notable findings here]

Keep the compiled report concise and actionable. The user wants to know what to fix, not read an essay about their essay.