ktg-plugin-marketplace/plugins/human-friendly-style/README.md

Human-Friendly Output Style for Claude Code

A shared output style that gives every plugin in this marketplace a consistent, plain-language tone. Install it once, activate it via /config, and Claude Code starts explaining work the way a person would — not the way a console dump does.

Solo-maintained, fork-and-own. This plugin is a starting point, not a vendor product. Issues are welcome as signals; pull requests are not accepted. See GOVERNANCE.md for the full model and what upstream provides.

AI-generated: all code produced by Claude Code through dialog-driven development. Full disclosure →

A Claude Code plugin that ships a single output style used across the ktg-plugin-marketplace. The style modifies Claude Code's system prompt so responses default to prose instead of bullet lists, hide noisy details (long file paths, raw shell commands, JSON blobs, stack traces) until the user asks for them, match the user's language, and stay honest about uncertainty. Claude Code's built-in coding instructions stay intact (keep-coding-instructions: true), so testing discipline, careful edits, and verification still apply — only the way Claude talks about the work changes.

---

The problem

Default Claude Code output is engineering output. Long absolute paths. Raw git invocations. JSON dumps. Stack traces. Bullet lists for everything. That is fine for a developer running terminal commands — it is the lingua franca of CLI work. It is also the wrong register when:

• A non-engineer is reading along over your shoulder
• You are documenting a session for someone who needs to understand what happened, not how it happened
• You are using Claude Code for non-code work (writing, research, planning) and the surrounding noise gets in the way
• You want each plugin in your toolkit to feel like part of one assistant, not eight different consoles

The other plugins in this marketplace cover specific domains (security, configuration, OKRs, Microsoft architecture, LinkedIn content, planning). Each has its own slash commands, agents, and hooks. What they did not have until now was a shared conversational tone — so installing two plugins meant getting two slightly different experiences glued together.

This plugin solves that with one short Markdown file and the official Claude Code plugin discovery mechanism.

---

What it does

The style file at output-styles/human-friendly.md declares eight directives that ride on top of Claude Code's normal system prompt:

#	Directive	What it changes
1	Explain what and why, not how	Claude describes the outcome and the reason; the technical mechanics stay reserved for when you ask
2	Use everyday language	Jargon gets translated; technical terms get one short clarification on first use
3	Hide noise by default	Long paths, raw commands, JSON, stack traces, and verbose tool output are summarized until you ask for the raw form
4	Summarize progress in prose	"I've read the configuration and found two settings that look outdated" beats a bullet list of every file inspected
5	Match the user's language	Norwegian when you write Norwegian, English otherwise — consistent within the conversation
6	Honest about uncertainty	"I think this should work, but I haven't tested it yet" beats fabricated confidence
7	Minimal formatting	Prose first; lists, tables, and headers only when structure genuinely helps comprehension
8	Failure framed plainly	Errors are described in human terms, with what is being done about them — no drama, no jargon walls

keep-coding-instructions: true in the frontmatter means Claude Code's built-in software-engineering instructions (test-first, secure code, careful edits, verification) stay active. The style only changes how Claude communicates about the work — not how the work itself is done.

---

Before / after

Same task, two styles, both real:

Default style — typical developer output:

I'll update the config. Let me first check the current state:

✓ Read /Users/you/project/.claude/settings.json
✓ Found 3 hooks: PreToolUse(Bash), PostToolUse(Edit), Stop

Plan:
- Add SessionStart hook
- Configure path: /Users/you/.claude/scripts/welcome.sh
- Set timeout: 5000ms

Proceeding with edit...

✓ Modified /Users/you/project/.claude/settings.json
   { "hooks": { "SessionStart": [{ "type": "command", ... }] } }

Done. Changes verified.

Human-Friendly style — same task, narrated:

I'll add a session-start welcome script to your project settings. Right now
you have three hooks set up; this adds a fourth that runs whenever you open
a session in this project. The script gets a five-second budget — long
enough to do something useful, short enough that it can't stall startup.

Done. The settings file is updated. If you want to see the exact JSON I
wrote, say the word.

Same operation, same correctness, different relationship to the reader. The information is still there — just folded behind a question if you want it.

---

Quick start

Install

Add the marketplace once, then install the plugin:

claude plugin marketplace add https://git.fromaitochitta.com/open/ktg-plugin-marketplace.git

In Claude Code:

/plugin install human-friendly-style@ktg-plugin-marketplace

Or enable directly in ~/.claude/settings.json:

{
  "enabledPlugins": {
    "human-friendly-style@ktg-plugin-marketplace": true
  }
}

Activate

/config

Choose Output style → Human-Friendly.

Output styles take effect from the next session — Claude Code holds the system prompt stable within a conversation for cache efficiency. Either start a new conversation, restart Claude Code, or use /clear to see the change.

Verify it took

A simple test: send a short prompt in Norwegian and one in English. The Norwegian prompt should get a Norwegian response; the English one should get English. If both come back in English, the style is not active yet — restart and try again.

---

How it works

+-----------------------------------------------------------+
|                    Claude Code Session                    |
|                                                           |
|  +-----------------+         +-----------------------+    |
|  | output-styles/  |  read   |   System prompt       |    |
|  | human-          | ----->  |   composition         |    |
|  | friendly.md     |  once   |                       |    |
|  +-----------------+         |  default coding       |    |
|                              |  instructions         |    |
|                              |    +                  |    |
|                              |  human-friendly       |    |
|                              |  directives           |    |
|                              +-----------+-----------+    |
|                                          |                |
|                              +-----------v-----------+    |
|                              |   Claude responds     |    |
|                              |   in plain language   |    |
|                              +-----------------------+    |
+-----------------------------------------------------------+

Claude Code auto-discovers .md files in the plugin's output-styles/ directory. When you select Human-Friendly in /config, the selection is persisted to settings (outputStyle: "Human-Friendly"). At the start of each session, Claude Code composes its system prompt by merging the default coding instructions with the directives in human-friendly.md. Same merge, same prompt, every turn — so prompt caching stays warm for the whole conversation.

Removing the style is as simple: /config → Output style → (default). The next session goes back to vanilla Claude Code output.

---

What this plugin does not do

By design, this plugin contains no commands, no agents, no hooks, no skills, no MCP servers. It ships exactly one file plus a manifest. That is the whole point — a shared conversational style should be a small, predictable thing that adds zero runtime overhead and zero surface area to audit.

If you need:

• Persona shaping (Cosmo Skyberg / specific advisor voices) — see ms-ai-architect
• Behavioral overrides (anti-sycophancy, reinforcement-loop detection) — see ai-psychosis
• Voice training (your own LinkedIn writing style) — see linkedin-thought-leadership
• Domain-specific workflows (OKR, security audits, planning pipelines) — see the other plugins

Those are domain plugins. This one is the shared chassis underneath them.

---

Cross-plugin use

The other plugins in this marketplace are designed to feel right with this style. Install human-friendly-style once, select it as your default, and the consistency shows up everywhere — /config-audit posture, /security audit, /architect, /okr:skriv, /ultraplan-local, all of them. You do not need to repeat the activation per plugin.

The style is optional. Every plugin in the marketplace works without it. This one just makes the conversation feel more like dialog and less like a console transcript.

---

Compatibility

Requirement	Version
Claude Code	recent versions with output style support
Platform	macOS, Linux, Windows
Network	None — output styles are local Markdown files

Output styles are a first-class feature in Claude Code's plugin system. Older Claude Code releases without /config → Output style support will install the plugin without errors but will not apply the style. See Claude Code Output Styles documentation for the canonical reference.

---

Versioning and stability

• Semantic versioning. Style content changes that affect Claude's output are minor or major bumps; pure typo fixes in the README are patch.
• The plugin is deliberately small and stable. The most likely future change is no change — the style is meant to be a quiet utility, not a feature roadmap.
• Variants (more code-focused, more terse, etc.) are out of scope for v1.x. If they ever ship, they would be additional .md files in the same output-styles/ directory.

---

License

MIT. Fork it, modify it, ship your own version under your own name.