<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo" -->

---
title: Claude Mods turn the agent harness to face humans
description: Claude Code now supports Claude Mods, TypeScript or JavaScript plugins that hook into prompts, tool calls, and rendering to reshape the agent's interface....
canonical: https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Claude Mods turn the agent harness to face humans | daily.dev
og:description: Claude Code now supports Claude Mods, TypeScript or JavaScript plugins that hook into prompts, tool calls, and rendering to reshape the agent's interface....
og:url: https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo
og:image: https://api.daily.dev/og/posts/0C3zymWxo.png
og:image:alt: Claude Mods turn the agent harness to face humans
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude Mods turn the agent harness to face humans

**[Trends](https://daily.dev/sources/trends)** · 3 min read · 5 upvotes · 4 comments

## Summary

Claude Code now supports Claude Mods, TypeScript or JavaScript plugins that hook into prompts, tool calls, and rendering to reshape the agent's interface. Anthropic ships six built-in mods, including 'You Should Know', which flagged uncommitted work before a demo edit would have overwritten it. A custom 'Next Steps Supervisor' mod fired on 99 of 111 turns since October 15, costing roughly 5 to 6k tokens per check on a 50k-token session. Another example, sf-headless-band, pins the current Salesforce default org above the prompt and shows production in red. Mods run unsandboxed with full file and prompt access, which the piece flags as a trust risk since the ecosystem is only days old.

## Content

Claude Code just became a platform for people who want to rearrange the furniture. Claude Mods are TypeScript or JavaScript plugins that hook into prompts, tool calls, and rendering, and the reaction so far is a mix of serious harness theory and people building arcade cabinets.

## The big-think camp

One writer calls it an "Attention Interface" and argues the mod system flips the agent harness to face the human instead of the machine. Anthropic ships six built-in mods. The one getting credit is "You Should Know", which flagged uncommitted work before a demo edit could overwrite it. The same author's custom "Next Steps Supervisor" fired on 99 of 111 turns since October 15, at an estimated 5 to 6k tokens per check on a 50k-token session. That's not free. They also line it up against OpenAI's Intelligent UI and Cursor's PR Routing & Approval.

@swyx is going bigger. His take: LLMs went from specific to foundation to specific again, and Claude Code is now "a malleable, general purpose hardened harness" that has "pretrained" on "quadrillions of tokens of learning" you can repurpose for any domain. Latent Space did a deep dive on the mods and picked favorites.

## The pragmatists

@avive spends 10 to 16 hours a day in Claude Code and Codex, and says walls of text wore him out by noon. His fix is action cards, one per code review issue, each with a recommended action, so decisions get made before implementation starts. The post drew enough interest that he published a gist.

On the practical end, a Salesforce walkthrough builds sf-headless-band, a mod that shows the default org above the prompt, with production in red. It also tracks which orgs the Salesforce MCP servers are hitting. It reads CLI config files directly for speed and loads permanently via CLAUDE_CODE_PLUGIN_DIRS.

## The fun police are not invited

Then there's the useful-adjacent stuff. cc-arcade adds nine mini-games, including Snake, Tetris, and a simplified Doom, and pauses when Claude finishes. The verbs mod turns all 188 spinner verbs into pixel-art animations. Mindful Claude runs guided breathing like box breathing and 4-7-8 while the agent works. One reviewer says waiting is now "actually fun". None of it changes coding output.

## The catch

Mods run unsandboxed with full file and prompt access. Install a Doom mod and you've handed it your prompts.

Hooks are the older, stricter sibling. The reference lists 33 lifecycle events. Exit 2 blocks an action, and a PreToolUse deny still blocks even with --dangerously-skip-permissions. Unlike a CLAUDE.md instruction, hooks are deterministic. Mods are where the vibes live; hooks are where the guarantees are.

## Questions this post answers

### What are Claude Mods in Claude Code and what can they do?

Claude Mods are TypeScript or JavaScript plugins for Claude Code that hook into prompts, tool calls, and rendering, letting developers reshape the agent's interface. Anthropic ships six built-in mods, including 'You Should Know', which warns about uncommitted work before an edit could overwrite it. Custom mods can add persistent UI elements, like a Salesforce org indicator built with the ui.render hook.

_Developers customizing their agent workflow can track new harness features like this on daily.dev._

### Is it safe to install third-party Claude Code plugins?

Not fully, since Claude Mods run unsandboxed with full file and prompt access, meaning a plugin that can redraw the terminal can also read everything passing through it. The plugin ecosystem is only days old, raising concern about a marketplace of unsandboxed code handling sensitive prompts before any vetting or guardrails exist.

_Anyone weighing agent plugin risk can follow how Claude Code's ecosystem matures on daily.dev._

### How much token overhead does an AI agent supervisor mod add per session?

A custom 'Next Steps Supervisor' mod for Claude Code triggered on 99 of 111 turns since October 15, consuming roughly 5,000 to 6,000 tokens per check within a 50,000-token session, a meaningful recurring cost for continuous oversight.

_Teams budgeting token spend on always-on agent checks can compare real costs like this on daily.dev._

## Community discussion

Top comments from developers on daily.dev.

**@nerdalytics** · 5 upvotes

> I created a mod to calculate cache freshness and costs. Technically, my extension scans the session transcript for token model and cache details. My mod also queries the API pricing page to render the cache price and decides whether to show a "start fresh" button.
>
> I also scan responses to calculate tokens-per-second performance.
>
> That's already quite a lot of trust debt I'm asking for.
>
> Now that I've built it, I can see how risky mods are.

## Similar posts on daily.dev

- [How to build your first Claude Code mod \(no coding required\)](https://daily.dev/posts/how-to-build-your-first-claude-code-mod-no-coding-required--rogzsa0qz) · LogRocket · 0 upvotes · 0 comments
- [Claude Code can now be modded like a video game, and I couldn't stop tinkering](https://daily.dev/posts/claude-code-can-now-be-modded-like-a-video-game-and-i-couldn-t-stop-tinkering-dzon9seci) · XDA Developers · 4 upvotes · 0 comments
- [How Claude Code mods change the agent loop](https://daily.dev/posts/how-claude-code-mods-change-the-agent-loop-doqr1tzdo) · Flavio Copes · 8 upvotes · 1 comments

---

Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#architecture](https://daily.dev/tags/architecture), [#claude-code](https://daily.dev/tags/claude-code), [#salesforce](https://daily.dev/tags/salesforce), [#typescript](https://daily.dev/tags/typescript)

[View this post on daily.dev](https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Claude Mods turn the agent harness to face humans","url":"https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo"},"datePublished":"2026-10-21T20:25:21.886Z","dateModified":"2026-10-29T02:14:10.954Z","description":"Claude Code now supports Claude Mods, TypeScript or JavaScript plugins that hook into prompts, tool calls, and rendering to reshape the agent's interface....","image":"https://pbs.twimg.com/media/HVLldA6bEAAuyvy.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HVLldA6bEAAuyvy.jpg","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":4,"discussionUrl":"https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":4}],"keywords":"ai-agents,architecture,claude-code,salesforce,typescript","timeRequired":"PT3M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Claude Mods turn the agent harness to face humans"}]}
{"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo","comment":[{"@type":"Comment","text":"I created a mod to calculate cache freshness and costs. Technically, my extension scans the session transcript for token model and cache details. My mod also queries the API pricing page to render the cache price and decides whether to show a “start fresh” button.\nI also scan responses to calculate tokens-per-second performance.\nThat’s already quite a lot of trust debt I’m asking for.\nNow that I’ve built it, I can see how risky mods are.","datePublished":"2026-10-21T21:34:09.162Z","url":"https://daily.dev/posts/0C3zymWxo#c-O78isLNTH","author":{"@type":"Person","name":"nerdalytics","url":"https://daily.dev/nerdalytics","image":"https://avatars.githubusercontent.com/u/97166791?v=4"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5}}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/claude-mods-turn-the-agent-harness-to-face-humans-0c3zymwxo#faq","mainEntity":[{"@type":"Question","name":"What are Claude Mods in Claude Code and what can they do?","acceptedAnswer":{"@type":"Answer","text":"Claude Mods are TypeScript or JavaScript plugins for Claude Code that hook into prompts, tool calls, and rendering, letting developers reshape the agent's interface. Anthropic ships six built-in mods, including 'You Should Know', which warns about uncommitted work before an edit could overwrite it. Custom mods can add persistent UI elements, like a Salesforce org indicator built with the ui.render hook. Developers customizing their agent workflow can track new harness features like this on daily.dev."}},{"@type":"Question","name":"Is it safe to install third-party Claude Code plugins?","acceptedAnswer":{"@type":"Answer","text":"Not fully, since Claude Mods run unsandboxed with full file and prompt access, meaning a plugin that can redraw the terminal can also read everything passing through it. The plugin ecosystem is only days old, raising concern about a marketplace of unsandboxed code handling sensitive prompts before any vetting or guardrails exist. Anyone weighing agent plugin risk can follow how Claude Code's ecosystem matures on daily.dev."}},{"@type":"Question","name":"How much token overhead does an AI agent supervisor mod add per session?","acceptedAnswer":{"@type":"Answer","text":"A custom 'Next Steps Supervisor' mod for Claude Code triggered on 99 of 111 turns since October 15, consuming roughly 5,000 to 6,000 tokens per check within a 50,000-token session, a meaningful recurring cost for continuous oversight. Teams budgeting token spend on always-on agent checks can compare real costs like this on daily.dev."}}]}
```

