From claude-ecosystem
Audits Claude Code output styles (.md files) for Markdown format, YAML frontmatter, structure, naming, and compatibility across plugin, project, and user scopes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-ecosystem:audit-output-stylesopusThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Audit Claude Code output styles for quality, compliance, and usability.
Audit Claude Code output styles for quality, compliance, and usability.
Before auditing, initialize the environment:
.claude/temp/) exists.The output-customization skill provides authoritative validation guidance (auto-loaded when this command runs).
.md, not .json)| Argument | Description |
|---|---|
| (none) | Audit all discoverable output styles |
plugin | Only audit plugin output styles |
project | Only audit .claude/output-styles/*.md |
user | Only audit ~/.claude/output-styles/*.md |
all | Audit all scopes explicitly |
--force | Audit regardless of modification status |
--skip-validation | Skip finding validation (faster, but may include false positives) |
Check plugin directories (plugins/*/output-styles/*.md), project directory (.claude/output-styles/*.md), and user directory (~/.claude/output-styles/*.md on Unix, %USERPROFILE%\.claude\output-styles\ on Windows).
Warn about any .json files found (wrong format - output styles must be markdown).
Parse the scope selector (plugin, project, user, all) and --force flag from command arguments. Filter discovered styles to match the requested scope.
Display mode (SMART or FORCE), styles discovered, plugins with styles, and file list with scope.
For each style, spawn the output-style-auditor subagent with the following context:
Run subagents in parallel batches of 3-5.
Subagents write findings to .claude/temp/ as both JSON (for recovery/aggregation) and markdown (for human review). The main conversation thread collects results and updates audit logs using its Write/Edit tools.
Unless --skip-validation flag is present:
audit-finding-validator agent with:
project_root: The captured project root pathaudit_type: "output-style"audit_files: List of .claude/temp/audit-*-output-style-*.json file paths.claude/temp/audit-filtered-findings.jsonIf --skip-validation flag is present:
Report total audited by scope, results, and details table. List format warnings for .json files with remediation (convert to .md).
Include validation statistics (if validation was performed):
Output styles require YAML frontmatter with name, description, and optionally keep-coding-instructions.
Use kebab-case with .md extension (e.g., my-style.md).
| Platform | User Styles |
|---|---|
| Unix | ~/.claude/output-styles/ |
| Windows | %USERPROFILE%\.claude\output-styles\ |
All audit results are written to .claude/audit/output-styles.md.
Use /audit-log output-styles to view current audit status.
User: /audit-output-styles
Claude: Discovering output styles...
## Audit Plan
**Mode**: SMART
**Styles discovered**: 3
**Plugins with styles**: 2
1. [plugin:claude-ecosystem] concise-coder.md
2. [plugin:soft-skills] code-review-comment.md
3. [project] custom-style.md
[Spawns output-style-auditor subagents]
## Audit Complete
| Scope | Style | Result | Score |
| --- | --- | --- | --- |
| plugin | concise-coder | PASS | 100/100 |
| project | custom-style | PASS | 98/100 |
User: /audit-output-styles plugin
Claude: Auditing plugin output styles only...
npx claudepluginhub melodic-software/claude-code-plugins --plugin claude-ecosystemManages Claude Code output styles via delegation to docs-management: built-ins, customs, frontmatter, /output-style switching, creation, comparisons, and config.
Design Claude Code output styles: persona, tone, and behavioral rules that replace the default system prompt. Invoke whenever task involves any interaction with output styles — creating, editing, evaluating, or changing how Claude communicates.
Guide for creating Claude Code output styles and reusable response-format contracts. Use to standardize agent output or decide between output styles, skills, and CLAUDE.md.