Claude Code Advanced Features

Portfolio Summary

Field Value

Field	Value
PRJ ID	PRJ-claude-code-features
Owner	Evan Rosado
Priority	P0 — Force Multiplier
Status	Active Development (v2)
Goal	Exponential increase in knowledge, wisdom, productivity, skills
Platform	Claude Code CLI (Opus 4.6)
Config Home	`~/.claude/` (stowed from dots-quantum)
Project Config	`domus-captures/.claude/` (project-scoped)
CLAUDE.md	873 lines (global behavioral instructions)
Settings	252 lines (107 allowed commands, 10 denied)
Skills	3 implemented (`/deploy`, `/worklog`, `/session`)
Hooks	5 event hooks (backup, validation, security)
Agents	10 total (4 global + 6 project-scoped)
Rules	5 path-specific (AsciiDoc, Lua/Neovim, Bash/Shell, Python, D2)
Output Styles	1 (professional)
Plugins	rust-analyzer-lsp

PRJ ID

PRJ-claude-code-features

Owner

Evan Rosado

Priority

P0 — Force Multiplier

Status

Active Development (v2)

Goal

Exponential increase in knowledge, wisdom, productivity, skills

Platform

Claude Code CLI (Opus 4.6)

Config Home

~/.claude/ (stowed from dots-quantum)

Project Config

domus-captures/.claude/ (project-scoped)

CLAUDE.md

873 lines (global behavioral instructions)

Settings

252 lines (107 allowed commands, 10 denied)

Skills

3 implemented (/deploy, /worklog, /session)

Hooks

5 event hooks (backup, validation, security)

Agents

10 total (4 global + 6 project-scoped)

Rules

5 path-specific (AsciiDoc, Lua/Neovim, Bash/Shell, Python, D2)

Output Styles

1 (professional)

Plugins

rust-analyzer-lsp

Why This Matters

You’re paying Anthropic instead of a bootcamp. Every feature implemented should:

Accelerate learning — Not just do work, but build muscle memory
Reduce friction — Automate repetitive workflows
Surface knowledge — Connect context across sessions
Challenge growth — Generate drills, track progress
Teach others — Every pattern documented is a transferable lesson

Architecture Overview

Figure 1. Claude Code Configuration Architecture (v2)

Deployment Model

All Claude Code configuration lives in dots-quantum (the stow-managed dotfiles repo) and is symlinked to ~/.claude/ via GNU Stow. This gives you:

Git-versioned config — every change tracked, diffable, reversible
Multi-machine portability — clone + stow = identical environment
Separation of concerns — global config (stowed) vs project config (repo-scoped)

Symlink Architecture

dots-quantum/claude/.claude/          # Canonical source (git-tracked)
├── CLAUDE.md                         # Global instructions (845 lines)
├── settings.json                     # Permissions, env, hooks (240 lines)
├── keybindings.json                  # Keyboard shortcuts
├── hooks/                            # Lifecycle scripts
│   ├── validate-asciidoc-attrs.sh    # PostToolUse: attribute validation
│   └── test-validate-asciidoc-attrs.sh
├── skills/                           # Reusable slash commands
│   ├── deploy/SKILL.md               # /deploy — push + rebuild
│   └── worklog/SKILL.md              # /worklog — daily log creation
├── agents/                           # Global subagents
│   └── doc-auditor.md                # Documentation staleness auditor
├── rules/                            # Path-specific instructions
│   ├── asciidoc.md                   # Auto-loads for *.adoc
│   └── lua-nvim.md                   # Auto-loads for nvim Lua
└── output-styles/                    # Response formatting
    └── professional.md               # Senior-engineer tone

~/.claude/                            # Symlinked destination (stowed)
├── CLAUDE.md        → dots-quantum/claude/.claude/CLAUDE.md
├── settings.json    → dots-quantum/claude/.claude/settings.json
├── keybindings.json → dots-quantum/claude/.claude/keybindings.json
├── hooks/           → dots-quantum/claude/.claude/hooks/
├── skills/          → dots-quantum/claude/.claude/skills/
├── agents/          → dots-quantum/claude/.claude/agents/
├── rules/           → dots-quantum/claude/.claude/rules/
├── output-styles/   → dots-quantum/claude/.claude/output-styles/
├── credentials.json → _vaults/mounted/credentials/claude/credentials.json
├── backups/                          # Auto-generated by hooks
├── history.jsonl                     # Conversation history
└── [runtime: cache/, plans/, tasks/, todos/, session-env/]

Project-Scoped Configuration

Repos can override global config with .claude/ at the repo root:

domus-captures/.claude/               # Project-scoped overrides
├── CLAUDE.md                         # Project context (442 lines)
├── settings.local.json               # Permission overrides
└── agents/                           # Project-specific agents
    ├── worklog-creator.md
    ├── build-fixer.md
    ├── carryover-updater.md
    ├── adoc-linter.md
    └── project-modularizer.md

Resolution order: Project CLAUDE.md supplements global. Project settings.local.json merges with global settings.json. Project agents add to (not replace) global agents.

Configuration Hierarchy

Layer Purpose Enforcement Location

Layer	Purpose	Enforcement	Location
CLAUDE.md	Behavioral guidelines, role definition, standards	Guidance (not hard-enforced)	`~/.claude/` + repo root
Skills	Reusable workflows invoked via `/command`	On-demand, user-triggered	`~/.claude/skills/`
Hooks	Auto-run shell scripts on lifecycle events	Deterministic execution, can block actions	`settings.json` → `~/.claude/hooks/`
Rules	Path-specific instructions for file types	Auto-loaded when editing matching files	`~/.claude/rules/`
Agents	Specialized subprocesses for complex tasks	Auto-dispatched by task description	`~/.claude/agents/` + `.claude/agents/`
Output Styles	Response formatting and tone	Applied globally per session	`~/.claude/output-styles/`
Settings	Permissions, sandbox, environment variables	Hard enforcement (allow/deny lists)	`settings.json` + `settings.local.json`

CLAUDE.md

Behavioral guidelines, role definition, standards

Guidance (not hard-enforced)

~/.claude/ + repo root

Skills

Reusable workflows invoked via /command

On-demand, user-triggered

~/.claude/skills/

Hooks

Auto-run shell scripts on lifecycle events

Deterministic execution, can block actions

settings.json → ~/.claude/hooks/

Rules

Path-specific instructions for file types

Auto-loaded when editing matching files

~/.claude/rules/

Agents

Specialized subprocesses for complex tasks

Auto-dispatched by task description

~/.claude/agents/ + .claude/agents/

Output Styles

Response formatting and tone

Applied globally per session

~/.claude/output-styles/

Settings

Permissions, sandbox, environment variables

Hard enforcement (allow/deny lists)

settings.json + settings.local.json

Improvement Proposals

Proposals from ecosystem audit — 2026-04-04. For team review and prioritization.

Priority	Proposal	Rationale	Effort
P1	Add field-notes partial (decisions behind skill/hook/agent choices)	17 partials document what is configured but not why. Decision rationale is lost between sessions.	M
P1	Document hook execution flow with diagram	Hooks fire on tool calls, prompt submission, and notifications. The interaction between pre/post hooks, deny rules, and shell execution is non-obvious.	M
P2	Troubleshooting guide for common hook failures	Hook failures silently block operations. A lookup table of error patterns and resolutions would save debugging time.	S
P2	Settings.json permission reference	97 allow rules and 10 deny rules are configured. A categorized table showing what each permission enables/blocks improves auditability.	M
P3	Agent capability matrix	Multiple agents (doc-auditor, adoc-linter, build-fixer, project-modularizer) exist globally and per-project. A matrix showing scope, tools, and use cases prevents duplication.	S

Priority

Proposal

Rationale

Effort

Add field-notes partial (decisions behind skill/hook/agent choices)

17 partials document what is configured but not why. Decision rationale is lost between sessions.

Document hook execution flow with diagram

Hooks fire on tool calls, prompt submission, and notifications. The interaction between pre/post hooks, deny rules, and shell execution is non-obvious.

Troubleshooting guide for common hook failures

Hook failures silently block operations. A lookup table of error patterns and resolutions would save debugging time.

Settings.json permission reference

97 allow rules and 10 deny rules are configured. A categorized table showing what each permission enables/blocks improves auditability.

Agent capability matrix

Multiple agents (doc-auditor, adoc-linter, build-fixer, project-modularizer) exist globally and per-project. A matrix showing scope, tools, and use cases prevents duplication.

Hook Execution Flow

A diagram showing the lifecycle of a Claude Code interaction:

User prompt → pre-submit hook → Claude response
  → tool call → pre-tool hook → [allow/deny check] → execution
  → post-tool hook → response continues
  → post-response hook → notification hooks

This should be a Mermaid sequence diagram in a partial, included from the architecture page.

Metadata

Field	Value
PRJ ID	PRJ-claude-code-features
Author	Evan Rosado
Created	2026-03-25
Last Updated	2026-04-03
Status	Active Development (v2)
Next Review	2026-04-15

Field

Value

PRJ ID

PRJ-claude-code-features

Author

Evan Rosado

Created

2026-03-25

Last Updated

2026-04-03

Status

Active Development (v2)

Next Review

2026-04-15

"The goal is not automation — it’s amplification. Every feature should make me MORE capable, not less engaged."