SkillFoxx

Agent Guides skill

Thirteen skills and seven guides for Claude Code on skills, subagents, hooks and memory, checked against the July 2026 docs

  • 369 ★ on the repository
  • Last commit
  • Apache-2.0
Skill

Extended access

Why this level

  • The `playwright` skill drives a browser from the terminal through the `playwright_cli.sh` wrapper
  • The export script reads transcripts in `~/.claude/projects/`, which may contain keys and paths from your sessions
All reasons and checks

tokenbender/agent-guidesAuthor:tokenbender

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/agent-guides

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The `playwright` skill drives a browser from the terminal through the `playwright_cli.sh` wrapper
  • The export script reads transcripts in `~/.claude/projects/`, which may contain keys and paths from your sessions

This is third-party code. Review the repository files before installing.

What it does

Agent Guides is the tokenbender repository with two sets of materials. The first is seven Claude Code guides on skills and their frontmatter, custom commands, subagents with worktree isolation, hooks, memory with path-scoped rules, code search and conversation history search. The second is thirteen skills in the usual `SKILL.md` layout with reference files, templates and scripts. The `claude-skills/` folder holds `multi-mind` (4-6 subagents argue in rounds), `capability-horizon-estimator` (a script that uses METR data to estimate whether a model can handle a task), `epistemic-libido`, `audit-sft-data-quality`, `tokenbending`, `worklog`, `paper-ocr-notes-pipeline`, `pdf-ocr-feedback`, `technical-writing-workflow`, `blade-and-contract`, `playwright`, `manim-math-explainer` and `x-thread-reader`. The `scripts/extract-claude-session.py` script exports a session from `~/.claude/projects/` to Markdown or cleaned JSONL.

Who it is for. Claude Code users who write their own skills, hooks and subagents and want to check their practices against the current docs.

Good fit when

  • You are writing your own skill and need the full schema of frontmatter, context injection and `context: fork`
  • The question is contested, such as an architecture choice, and you want specialists with different positions: `multi-mind` does that
  • You want to estimate whether a model can handle a four-hour task using a success curve from METR data
  • You want to set up guardrail hooks: the guide collects five working recipes
  • You need a previous Claude Code session exported to Markdown

Not a fit when

  • You need a ready integration with a service such as GitHub or Slack: the collection holds methods and guides, with no connectors
  • Your working agent is not Claude Code: the guides describe its folders and mechanisms
  • You need a minimal set: the copy-all command enables thirteen skills at once, so take them one at a time

Example request

Use multi-mind to decide whether to split our monolith into services: three rounds, let the specialists argue

Limitations

The guides are tied to the Claude Code ecosystem as of July 2026: field names, commands and subagent modes may change, and the author updates the text by hand.

Some skills lean on outside pieces. `playwright` works through `playwright-cli` and the `playwright_cli.sh` wrapper, `x-thread-reader` reads threads through the twitter-thread.com service, and `capability-horizon-estimator` calculates from METR data in `references/horizon-data.md`.

`multi-mind` launches 4-6 subagents for three rounds by default, and that is a noticeable token spend.

How to disable. Delete the folder of the skill you want to remove from `~/.claude/skills`, for example `multi-mind`, and restart Claude Code. The guides install nothing.

FAQ

How do I install a single skill from the collection?

Clone the repository and run `cp -r claude-skills/multi-mind ~/.claude/skills/` with the name you need. `cp -R claude-skills/*/ ~/.claude/skills/` installs all thirteen at once.

What does multi-mind do?

It hands the question to 4-6 parallel subagents with different views, runs rounds of cross-critique (three by default) and merges the result. It suits contested decisions. For factual questions the skill text points to `/deep-research`.

How does capability-horizon-estimator work?

The `estimate.py` script takes the model horizon in minutes (`--h50`) and the task length (`--task-min`) and prints the single-attempt success probability, a verdict and a wall-clock estimate. For a 719-minute horizon and a 240-minute task the example gives 70.6 percent.

Which guides does the repository have?

Seven: on skills, custom commands, subagents, hooks, memory, code search and conversation search. Each one is a separate Markdown file in the repository root.

How do I export a Claude Code session?

The `scripts/extract-claude-session.py` script reads JSONL transcripts from `~/.claude/projects/` and writes the chosen session to Markdown or cleaned JSONL.

Discussion

No questions or reviews yet.

Skills

A Claude Code reference plus a collection of over 70 skills for development, security, data and docs, with agent configs

SkillBroad access4.7K

11 skills for Claude Code: SKILL.md authoring, prompt tuning, refactoring, change review and planning in files.

SkillMinimal access286

A working folder for Claude Code where the agent asks questions and writes a new skill by Anthropic's guidance, not from memory.

SkillExtended access242

A set of skills, slash commands and agents for Claude Code that helps you build your own skills, agents, hooks, slash commands and prompts

SkillExtended access899

Skills, mods, subagents and hooks for Claude Code: the agent inspects your setup and installs what fits after your approval.

SkillExtended access98

The author's personal Claude Code toolbox: meta-prompt and todo commands, skills to create skills, hooks and MCP servers, auditors and a Ralph loop

PluginBroad access2K

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialRepository starsMain difference
Agent Guides (this card)no369
Karpathy-inspired Claude Code guidelinesno232,688a short charter for CLAUDE.md and three skills, while here there are guides to Claude Code mechanisms and thirteen skills
Claude Skills for full-stack developersno12,342skills for languages, frameworks and DevOps, while here the skills cover working method with an agent, plus guides

README in short

It starts with a table of guides, each described, and a hint on which Claude Code mechanism to pick for a fact, a procedure, isolated work or a hard limit. Then comes a table of thirteen skills, a run example of the horizon estimator and a description of the session export script.

README badge

Original SKILL.md

---
description: Multi-specialist collaborative analysis — fans a topic out to parallel specialist subagents, cross-pollinates findings, and synthesizes across rounds
argument-hint: "<topic> [rounds=3]"
disable-model-invocation: true
---

# Multi-Mind - Subagent-Based Collaborative Analysis

Execute a multi-specialist collaborative analysis on: **$ARGUMENTS**

Parse an optional `rounds=N` parameter (default 3).

## How this maps to Claude Code's agent system

- Launch each specialist as a **parallel subagent** via the Task tool — they run concurrently, each in its own context window, and return only their analysis.
- For one-off runs, inline subagent prompts (below) are enough. For topics you revisit, make the specialists persistent by writing definitions to `.claude/agents/<role>.md` (see the [Subagents Guide](../claude-subagents-guide.md)) — Claude will then route to them by description automatically.
- Subagents run in the background by default; launch all specialists for a round in a single message so they execute in parallel.
- If the question needs codebase-wide changes (not just analysis), the bundled `/batch` skill is a better fit. For long-running collaborative sessions, see agent teams in the Claude Code docs.

**When to use this vs. `/deep-research`:** the bundled `/deep-research` workflow fans out web searches and synthesizes a cited report — use it for factual questions. Use multi-mind for *judgment* questions (architecture tradeoffs, strategy, design decisions) where decorrelated expert perspectives and adversarial cross-pollination matter more than citation coverage.

## Phase 1: Specialist Assignment & Research

Analyze the topic and determine 4-6 specialist roles with maximally decorrelated perspectives. Then launch them all in parallel, e.g.:

```
Task tool, subagent_type: general-purpose — run all in one block:

1. "You are a technical specialist. Research [topic] focusing on implementation
   details, architecture, performance, and technical risk. Use web search for
   current documentation and case studies. Return ≤500 words of findings."

2. "You are a business strategy specialist. Analyze [topic] from market dynamics,
   competitive landscape, ROI, and strategic positioning. Use web search for
   current market reports. Return ≤500 words."

3. "You are a user experience / adoption specialist. Investigate [topic] for
   user needs, usability concerns, and adoption barriers. Return ≤500 words."

[... additional specialists as the topic demands — security, regulatory,
 ethics, operations, contrarian/red-team ...]
```

**Specialist selection criteria**:
- Unique domain expertise relevant to the topic

Report a problem