Agent-Skills.md

Agent Skills: List Claude Code Components

List Claude Code components. Actions: env-vars, sessions, settings, skills, tools.

UncategorizedID: melodic-software/claude-code-plugins/list

Repository

melodic-software/claude-code-plugins
melodic-softwareLicense: MIT
417

Install this agent skill to your local

pnpm dlx add-skill https://github.com/melodic-software/claude-code-plugins/tree/HEAD/plugins/claude-ecosystem/skills/list

Skill Files

Browse the full folder contents for list.

Download Skill

Loading file tree…

plugins/claude-ecosystem/skills/list/SKILL.md

Skill Metadata

Name
list
Description
"List Claude Code components. Actions: env-vars, sessions, settings, skills, tools."

List Claude Code Components

List various Claude Code components based on the requested action.

Argument Routing

| Action | Description | |--------|-------------| | skills (default) | List all available skills with descriptions | | env-vars | List environment variables (configured or researched) | | sessions | List recent sessions with sizes, ages, resumability | | settings | List settings.json configuration options | | tools | List core Claude Code tools with parameters |

When invoked without arguments, default to skills.

Parse $ARGUMENTS to determine the action. The first token is the action keyword. Remaining tokens are passed as sub-arguments to the action handler.

Action: skills

List ALL available Claude Code skills by running the list_skills.py script.

Step 1: Find the Script

Locate list_skills.py in the claude-ecosystem plugin:

# Check installed plugins first
find ~/.claude/plugins -name "list_skills.py" -path "*claude-ecosystem*" 2>/dev/null | head -1

If not found, check current repo (for development):

find . -name "list_skills.py" -path "*claude-ecosystem*" 2>/dev/null | head -1

Step 2: Run the Script

Execute with Python using the path found in Step 1:

python "<found-path>/list_skills.py"

What the Script Scans

The script finds ALL skills from:

  • Personal skills: ~/.claude/skills/*/SKILL.md
  • Project skills: .claude/skills/*/SKILL.md (current working directory)
  • Plugin skills: ~/.claude/plugins/marketplaces/*/*/skills/*/SKILL.md

Output Format

The script outputs formatted markdown:

## Personal Skills (~/.claude/skills/)
### **skill-name**
Description from SKILL.md frontmatter.
---

## Project Skills (.claude/skills/)
### **skill-name**
Description from SKILL.md frontmatter.
---

## Plugin Skills
### **plugin-name:skill-name**
Description from SKILL.md frontmatter.
---

**Total: X skills** (Y personal, Z project, W plugin)

Output the script results directly - they are already formatted.

Action: env-vars

Research and list ALL known Claude Code environment variables from multiple sources, or show currently configured variables across all settings scopes.

Sub-arguments

| Argument | Description | |----------|-------------| | --configured | Show currently configured env vars from all settings files | | --official-only | Only show variables documented in official docs | | --changelog-only | Only extract from CHANGELOG.md | | --full-research | Run comprehensive MCP research (slower, more complete) | | (none) | Default: official docs + changelog extraction |

Behavior Matrix

| Arguments | Behavior | |-----------|----------| | (none) | Research: official docs + changelog | | --configured | Read current settings files only (no research) | | --official-only | Research: official docs only | | --changelog-only | Research: changelog extraction only | | --full-research | Research: all sources including MCP agents |

Mode: --configured

When --configured is passed, skip all research and read currently configured environment variables from all settings scopes.

Settings File Hierarchy (Precedence: Enterprise < User < Project < Local)

| Scope | Path | Description | |-------|------|-------------| | Enterprise | ~/.claude/enterprise/settings.json | IT-managed defaults | | User/Global | ~/.claude/settings.json | User preferences | | Project | .claude/settings.json | Project-specific settings | | Local | .claude/settings.local.json | Gitignored local overrides |

Workflow

  1. Read each settings file if it exists
  2. Extract env section from each
  3. Calculate effective values (later scopes override earlier)
  4. Display merged view with source indication

Output Format

## Currently Configured Environment Variables

| Variable | Value | Source | Overridden By |
|----------|-------|--------|---------------|
| DISABLE_TELEMETRY | "1" | user | project |
| ANTHROPIC_API_KEY | "sk-***" | user | - |
| CLAUDE_HOOK_LOG_DIR | "./logs" | project | - |
| DISABLE_AUTOUPDATER | "0" | enterprise | - |

### By Scope

**Enterprise** (~/.claude/enterprise/settings.json)
- DISABLE_AUTOUPDATER = "0"

**User** (~/.claude/settings.json)
- DISABLE_TELEMETRY = "1"
- ANTHROPIC_API_KEY = "sk-***" (masked)

**Project** (.claude/settings.json)
- DISABLE_TELEMETRY = "0" (overrides user)
- CLAUDE_HOOK_LOG_DIR = "./logs"

**Local** (.claude/settings.local.json)
(none configured)

Security Note

Mask sensitive values in output:

  • ANTHROPIC_API_KEY -> "sk-***" (show prefix only)
  • AWS_* credentials -> masked
  • GOOGLE_* credentials -> masked

Mode: Research (Default)

When no --configured flag, research environment variables from multiple sources.

Step 1: Parallel Research via Agents

Launch research agents in parallel to gather env var information from multiple sources.

Agent 1: MCP Research (full-research mode only)

Spawn mcp-research agent:

Research Claude Code environment variables using perplexity and firecrawl.

Search for:
- "Claude Code environment variables configuration"
- "anthropic claude cli env vars"
- "ANTHROPIC_API_KEY CLAUDE_CODE environment"

Extract ALL environment variables mentioned, including:
- Variable name
- Default value (if mentioned)
- Purpose/description
- Source URL

Return structured findings with source citations.
Agent 2: Official Documentation

Spawn claude-code-guide agent:

First WebFetch https://code.claude.com/docs/en/claude_code_docs_map.md to find
relevant doc pages about environment variables and settings.

Then WebFetch these specific pages:
- Settings/configuration pages
- Environment variables documentation
- Any pages mentioning ANTHROPIC_*, CLAUDE_*, DISABLE_*, ENABLE_*

Do NOT use Skill tool (not available).
Return complete env var tables with source URLs.
Skill: Local Documentation Cache

Invoke docs-management skill with query:

Search for "environment variables" and "env" in Claude Code documentation.
Include any settings, configuration, or customization topics.

Step 2: Changelog Extraction

WebFetch the CHANGELOG.md to extract environment variables:

URL: https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md

Extract all patterns matching:

  • ANTHROPIC_*
  • CLAUDE_*
  • DISABLE_*
  • ENABLE_*
  • *_TIMEOUT
  • *_API_KEY
  • HTTP_PROXY, HTTPS_PROXY, NO_PROXY

Note the version where each variable was introduced.

Step 3: Consolidate and Categorize

Merge findings from all sources, deduplicate, and categorize:

| Category | Criteria | |----------|----------| | OFFICIAL | Documented on code.claude.com/docs | | CHANGELOG | In CHANGELOG but not in official docs | | DISCOVERED | Found via MCP research only |

Step 4: Output Report

Format the complete list as structured tables:

## Claude Code Environment Variables

### Official (Documented)
| Variable | Default | Purpose | Category |
|----------|---------|---------|----------|
| ANTHROPIC_API_KEY | (required) | API key for Anthropic API | Auth |
| CLAUDE_CODE_USE_BEDROCK | "0" | Use AWS Bedrock instead of Anthropic API | Provider |
| CLAUDE_CODE_USE_VERTEX | "0" | Use Google Vertex AI | Provider |
| ... | ... | ... | ... |

### Changelog (Undocumented)
| Variable | Default | Purpose | Version Added |
|----------|---------|---------|---------------|
| CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR | "0" | Keep project working directory | v2.x |
| ... | ... | ... | ... |

### Discovered (MCP Research)
| Variable | Default | Purpose | Source |
|----------|---------|---------|--------|
| ... | ... | ... | [source URL] |

---

## Sources
- Official: https://code.claude.com/docs/en/settings
- Changelog: https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
- MCP: perplexity, firecrawl (if --full-research)

Known Environment Variables (Quick Reference)

These are commonly used variables (verify against current docs):

Authentication & Provider

  • ANTHROPIC_API_KEY - API key (required unless using Bedrock/Vertex)
  • CLAUDE_CODE_USE_BEDROCK - Use AWS Bedrock
  • CLAUDE_CODE_USE_VERTEX - Use Google Vertex AI

Proxy & Network

  • HTTP_PROXY / HTTPS_PROXY - Proxy configuration
  • NO_PROXY - Proxy bypass list

Behavior

  • CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR - Working directory behavior
  • DISABLE_AUTOUPDATER - Disable auto-updates
  • DISABLE_TELEMETRY - Disable telemetry

Hooks

  • CLAUDE_HOOK_* - Custom hook environment variables (repo-specific)

Note: This quick reference may be outdated. Always run the full command for current information.

Action: sessions

Display recent Claude Code sessions for the current project (or all projects), showing size, age, and whether they can be resumed.

Sub-arguments

| Argument | Description | Default | |----------|-------------|---------| | --all | List sessions from all projects | false | | --limit N | Maximum sessions to display | 10 | | --agents | Include agent transcripts in listing | false |

Step 1: Determine Target Directory

if ALL_PROJECTS; then
  TARGET_DIR="$HOME/.claude/projects"
else
  PROJECT_PATH=$(pwd | sed 's/[\/:]/-/g' | sed 's/^-//')
  TARGET_DIR="$HOME/.claude/projects/$PROJECT_PATH"
fi

Step 2: List Sessions

# Find session files (not agent files), sorted by modification time
find "$TARGET_DIR" -name "*.jsonl" ! -name "agent-*" -type f -printf '%T@ %s %p\n' 2>/dev/null | \
  sort -rn | head -$LIMIT | while read timestamp size path; do
    filename=$(basename "$path")
    age=$(( ($(date +%s) - ${timestamp%.*}) / 86400 ))
    size_h=$(numfmt --to=iec-i --suffix=B $size)
    echo "$filename  $size_h  ${age}d ago"
done

Step 3: Display Results

Recent Sessions
===============
Project: D--repos-gh-melodic-claude-code-plugins

 #  Session ID                              Size    Age     Resumable
--- --------------------------------------- ------- ------- ---------
 1  4165caab-4f86-4fc8-a901-fdcf2307a7cd   8.4M    today   Yes
 2  faf4b0f3-1f1f-438a-b40b-1a7ef4cb1718   7.9M    today   Yes
 3  55e42424-f5b6-4a26-98d5-1bed9176bf50   15M     1d      Yes
 4  336198a5-dfa4-4266-86dc-3d90158b8e30   15M     2d      Yes
 5  83d6bc93-3337-4b9e-8818-7141effbbcf0   14M     4d      Yes
...

Total: 507 sessions (645M)
Tip: Use /cleanup-sessions 7 to remove old sessions

Notes

  • Sessions can be resumed using claude --resume {session-id}
  • Session IDs are UUIDs assigned by Claude Code
  • Large sessions (>10MB) indicate long conversations
  • Use /session-stats for aggregate statistics

Action: settings

Research and list all known Claude Code settings.json configuration options, show the current settings hierarchy with precedence, or display only current project settings.

Note: Environment variables are documented separately. Use the env-vars action for environment variable documentation.

Sub-arguments

| Argument | Description | |----------|-------------| | (none) | Default: Full hierarchy - show user/project/local/managed settings with precedence | | --fields-only | Schema reference - list all available fields with types and defaults | | --current | Show current project settings values only (no hierarchy) | | --category <name> | Filter by category: core, permissions, sandbox, hooks, plugins, attribution | | --scope <scope> | Limit to specific scope: user, project, local, managed |

Behavior Matrix

| Arguments | Behavior | |-----------|----------| | (none) | Read all settings files, display hierarchy with precedence | | --fields-only | Research schema, display available fields with types/defaults | | --current | Read project settings only, display explicit values | | --category X | Filter output to specific category (works with all modes) | | --scope X | Limit hierarchy to specific scope (hierarchy mode only) |

Settings File Hierarchy

Claude Code settings follow a precedence order (higher overrides lower):

| Priority | Scope | Path | Description | |----------|-------|------|-------------| | 1 (Highest) | Managed | ~/.claude/managed/settings.json | IT-enforced (cannot be overridden) | | 2 | CLI Args | --model, etc. | Command line overrides | | 3 | Local | .claude/settings.local.json | Gitignored local overrides | | 4 | Project | .claude/settings.json | Shared project settings | | 5 (Lowest) | User | ~/.claude/settings.json | User preferences |

Mode: Default (Full Hierarchy)

When no arguments are provided, display the complete settings hierarchy showing precedence.

Workflow

  1. Read all settings files that exist:

    • ~/.claude/managed/settings.json (managed)
    • ~/.claude/settings.json (user)
    • .claude/settings.json (project)
    • .claude/settings.local.json (local)

  2. Calculate effective values - Later scopes override earlier (except managed which is enforced)

  3. Display merged view with source indication

Output Format

## Settings Hierarchy (Effective Configuration)

### Precedence Order (highest to lowest)
1. Managed settings (enforced) - cannot be overridden
2. Command line arguments
3. Local project settings (.claude/settings.local.json)
4. Shared project settings (.claude/settings.json)
5. User settings (~/.claude/settings.json)

### Core Settings
| Setting | Effective Value | Source | Overridden? |
|---------|-----------------|--------|-------------|
| model | opus | project | - |
| cleanupPeriodDays | 30 | default | - |
| respectGitignore | true | default | - |
| plansDirectory | ~/.claude/plans | default | - |
| showTurnDuration | true | default | - |

### Permissions Settings
| Setting | Effective Value | Source | Overridden? |
|---------|-----------------|--------|-------------|
| defaultMode | bypassPermissions | project | user |
| allow | [list...] | project | - |
| deny | [list...] | project | - |

### Sandbox Settings
(Show if sandbox configured, otherwise note "Not configured - defaults to disabled")

### Hook Settings
| Setting | Effective Value | Source | Overridden? |
|---------|-----------------|--------|-------------|
| disableAllHooks | false | project | - |
| hooks | {} | default | - |

### Status Line
(Show statusLine configuration if set)

### Plugin Settings
(Show enabledPlugins if configured)

---

## Environment Variables
Use `list env-vars` action for environment variable documentation.

Mode: --fields-only (Schema Reference)

When --fields-only is passed, research and display all available settings fields with types and defaults.

Workflow

  1. Research schema via docs-management skill and MCP servers
  2. Extract all known fields from official docs, changelog, and schema
  3. Categorize and display with types, defaults, and version info

Research Sources

Source 1: docs-management Skill

Invoke docs-management skill:

Search for "settings.json" and "configuration" in Claude Code documentation.
Include settings schema, available fields, and configuration options.
Source 2: MCP Research (comprehensive)

Spawn mcp-research agent:

Research Claude Code settings.json schema and configuration options using perplexity and firecrawl.

Search for:
- "Claude Code settings.json schema"
- "claude code configuration options"
- Site:code.claude.com settings

Extract ALL settings fields with:
- Field name
- Type (string, boolean, number, object, array)
- Default value
- Description
- Version introduced (if known)

Return structured findings with source citations.
Source 3: JSON Schema

Reference the schema at: https://json.schemastore.org/claude-code-settings.json

Output Format

## Available Settings Fields

### Core Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| $schema | string | - | JSON schema reference | - |
| model | string | "sonnet" | Default model alias or full ID | - |
| language | string | (system) | Interface language | v2.1.0 |
| autoUpdatesChannel | enum | "latest" | "stable" or "latest" | - |
| cleanupPeriodDays | number | 30 | Days before session cleanup | - |
| respectGitignore | boolean | true | Honor .gitignore patterns | - |
| outputStyle | string | null | Output formatting style | - |
| alwaysThinkingEnabled | boolean | false | Enable extended thinking | - |
| plansDirectory | string | ~/.claude/plans | Plan file storage | v2.1.9 |
| showTurnDuration | boolean | true | Show turn timing | v2.1.7 |

### Permissions Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| permissions.defaultMode | enum | "default" | default/acceptEdits/plan/dontAsk/bypassPermissions | - |
| permissions.allow | array | [] | Allowed tool patterns | - |
| permissions.deny | array | [] | Denied tool patterns | - |
| permissions.ask | array | [] | Always-ask tool patterns | - |
| permissions.additionalDirectories | array | [] | Extra allowed directories | - |
| permissions.disableBypassPermissionsMode | string | null | Set to "disable" to block --dangerously-skip-permissions | - |

### Sandbox Settings (macOS/Linux only)
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| sandbox.enabled | boolean | false | Enable sandboxed bash | - |
| sandbox.autoAllowBashIfSandboxed | boolean | true | Auto-allow bash in sandbox | - |
| sandbox.excludedCommands | array | [] | Commands bypassing sandbox | - |
| sandbox.allowUnsandboxedCommands | boolean | true | Allow unsandboxed commands | - |
| sandbox.enableWeakerNestedSandbox | boolean | false | Linux nested sandbox | - |
| sandbox.network.allowUnixSockets | array | [] | Allowed Unix sockets | - |
| sandbox.network.allowLocalBinding | boolean | false | Allow local port binding | - |
| sandbox.network.httpProxyPort | number | null | HTTP proxy port | - |
| sandbox.network.socksProxyPort | number | null | SOCKS proxy port | - |

### Hook Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| hooks | object | {} | Hook configuration by event | - |
| disableAllHooks | boolean | false | Disable all hooks globally | - |
| allowManagedHooksOnly | boolean | false | Only allow managed hooks | - |

### Attribution Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| attribution.commit | string | null | Custom commit attribution | - |
| attribution.pr | string | null | Custom PR attribution | - |

### Status Line Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| statusLine.type | enum | - | "command" for custom status | - |
| statusLine.command | string | - | Status line command | - |
| statusLine.padding | number | 0 | Status line padding | - |
| statusLine.timeout | number | 5000 | Command timeout (ms) | - |

### Plugin Settings
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| enabledPlugins | object | {} | Enabled plugins by namespace | - |
| extraKnownMarketplaces | object | {} | Additional plugin marketplaces | - |

### Environment Variables
| Field | Type | Default | Description | Since |
|-------|------|---------|-------------|-------|
| env | object | {} | Environment variables | - |

**Note:** For detailed environment variable documentation, use the `env-vars` action.

---

## Sources
- Schema: https://json.schemastore.org/claude-code-settings.json
- Docs: https://code.claude.com/docs/en/settings
- [MCP research citations]

Mode: --current (Current Project Only)

When --current is passed, display only the explicitly configured values in the current project's settings.

Workflow

  1. Read .claude/settings.json (project settings)
  2. Display only explicitly set values (not defaults)
  3. Skip research and hierarchy calculation

Output Format

## Current Project Settings

**File:** `.claude/settings.json`

### Explicitly Configured Values

#### Core Settings
(none explicitly set - using defaults)

#### Environment Variables
54 variables configured - see `.claude/settings.json` or use `list env-vars --configured`

#### Permissions
- defaultMode: "bypassPermissions"
- allow: 26 patterns configured
- deny: 24 patterns configured

#### Hooks
- disableAllHooks: false

#### Status Line
- type: "command"
- command: "bun x ccusage statusline ..."
- padding: 0
- timeout: 5000

---

Use `--fields-only` to see all available settings fields.
Use `list env-vars` for environment variable details.

Category Definitions

| Category | Fields Included | |----------|-----------------| | core | model, language, autoUpdatesChannel, cleanupPeriodDays, respectGitignore, outputStyle, alwaysThinkingEnabled, plansDirectory, showTurnDuration | | permissions | permissions.* (all permission fields) | | sandbox | sandbox.* (all sandbox fields) | | hooks | hooks, disableAllHooks, allowManagedHooksOnly | | plugins | enabledPlugins, extraKnownMarketplaces | | attribution | attribution.commit, attribution.pr |

Action: tools

Introspect your current tool context and enumerate all core Claude Code tools with their complete parameter schemas.

Scope

List only core built-in tools (not installed MCP server tools):

  • File Operations: Read, Write, Edit, Glob, Grep, NotebookEdit
  • Execution: Bash, Task, TaskOutput, KillShell
  • User Interaction: AskUserQuestion, TodoWrite
  • Planning & Workflow: EnterPlanMode, ExitPlanMode
  • Web & Network: WebFetch, WebSearch
  • Code Intelligence: LSP, Skill
  • MCP Resources: ListMcpResourcesTool, ReadMcpResourceTool

Sub-arguments

| Argument | Description | |----------|-------------| | --category <cat> | Filter by category: file, execution, interaction, planning, web, code, mcp | | --tool <name> | Show detailed info for a specific tool (e.g., --tool Read) |

Execution

Step 1: Parse Arguments

Parse any provided arguments:

  • If --category specified, note the category filter
  • If --tool specified, note the specific tool name

Step 2: Introspect Tool Context

Examine your current system context to enumerate all available core tools. For each tool, extract:

  1. Tool name - The function name
  2. Description - What the tool does (from the tool definition)
  3. Parameters - All parameters with:
    • Name
    • Type (string, number, boolean, array, object, enum)
    • Required or optional
    • Default value (if any)
    • Constraints (min, max, enum values)
    • Description

Step 3: Categorize

Group tools into these categories:

| Category | Tools | |----------|-------| | file | Read, Write, Edit, Glob, Grep, NotebookEdit | | execution | Bash, Task, TaskOutput, KillShell | | interaction | AskUserQuestion, TodoWrite | | planning | EnterPlanMode, ExitPlanMode | | web | WebFetch, WebSearch | | code | LSP, Skill | | mcp | ListMcpResourcesTool, ReadMcpResourceTool |

Step 4: Apply Filters

  • If --category was specified, show only tools in that category
  • If --tool was specified, show only that tool with full detailed view

Step 5: Output

Summary View (default or --category):

## Core Claude Code Tools (N tools)

### Category Name (N tools)

| Tool | Description |
|------|-------------|
| ToolName | Brief description |
| ... | ... |

Detailed View (--tool ToolName):

## ToolName

Full description from tool definition.

### Parameters

| Name | Type | Required | Default | Description |
|------|------|----------|---------|-------------|
| param_name | type | Yes/No | value | Description |

### Constraints

- List any enum values, min/max, format requirements

### Usage Notes

- Key behaviors
- Limitations
- Best practices