[FEATURE] Ability to run slash command in CLAUDE.md

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/2735
2. https://github.com/anthropics/claude-code/issues/10282
3. https://github.com/anthropics/claude-code/issues/10808

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Feature Analysis: Auto-Execute Slash Commands in CLAUDE.md

This is a well-thought-out feature request. Here's the technical approach and implementation path.

Problem Validation

Pain point confirmed: Users must manually run setup commands every session.

Current workarounds:

• ❌ Hooks can't run slash commands
• ❌ CLAUDE.md instructions ignored by Claude
• ✅ Manual reminder (current solution)

Proposed Solution: AutoExec Block

In CLAUDE.md:

# Project Configuration

<!--AUTOEXEC
/setup-claude
/skill load validation-expert
/skill load architecture-reviewer
-->

## Instructions
... rest of your file ...

Parsing logic:

// src/config/ClaudeMdParser.ts

interface AutoExecBlock {
  commands: string[];
  trigger: 'session_start' | 'context_clear' | 'both';
}

function parseClaudeMd(content: string): ClaudeMdConfig {
  const autoexecMatch = content.match(/<!--AUTOEXEC\n([\s\S]*?)\n-->/);
  
  if (!autoexecMatch) {
    return { autoexec: null };
  }
  
  const commands = autoexecMatch[1]
    .split('\n')
    .filter(line => line.trim().startsWith('/'))
    .map(line => line.trim());
  
  return {
    autoexec: {
      commands,
      trigger: 'both'  // Default: run on start and after /clear
    }
  };
}

Execution flow:

// src/session/SessionManager.ts

class SessionManager {
  async startSession() {
    // 1. Load CLAUDE.md
    const config = await loadClaudeMd();
    
    // 2. Execute autoexec commands
    if (config.autoexec) {
      for (const cmd of config.autoexec.commands) {
        await this.executeSlashCommand(cmd);
      }
    }
    
    // 3. Start normal session
    this.beginConversation();
  }
  
  async clearContext() {
    // When /clear is called
    this.messages = [];
    
    // Re-run autoexec
    const config = await loadClaudeMd();
    if (config.autoexec?.trigger !== 'session_start') {
      for (const cmd of config.autoexec.commands) {
        await this.executeSlashCommand(cmd);
      }
    }
  }
}

Security Considerations

Risk: Malicious CLAUDE.md could auto-execute harmful commands

Mitigation 1: Whitelist

// Only allow safe commands:
const SAFE_AUTOEXEC_COMMANDS = [
  '/skill',
  '/config',
  '/help',
  '/project'  
];

// Block dangerous ones:
const BLOCKED_COMMANDS = [
  '/exec',  // Arbitrary bash
  '/run',
  '/delete'
];

Mitigation 2: User Confirmation

Claude Code detected autoexec commands in CLAUDE.md:
  /setup-claude
  /skill load validation-expert

Execute these commands? [y/N]: _

Mitigation 3: Project Trust System

// src/security/ProjectTrust.ts

interface ProjectTrustLevel {
  path: string;
  trusted: boolean;
  autoexec_allowed: boolean;
}

// User must explicitly trust project:
$ claude --trust-project .
✓ Project trusted. CLAUDE.md autoexec will run automatically.

Alternative: Hook-Based Approach

Add new hook: SessionStart

# .claude/hooks/session-start.sh
#!/bin/bash

# This runs BEFORE first message
# Can execute slash commands via special syntax

echo "/setup-claude" >> $CLAUDE_COMMAND_QUEUE
echo "/skill load validator" >> $CLAUDE_COMMAND_QUEUE

Implementation:

// Queue commands from hook
const queue = readFile(process.env.CLAUDE_COMMAND_QUEUE);
for (const cmd of queue.split('\n')) {
  if (cmd.startsWith('/')) {
    await executeSlashCommand(cmd);
  }
}

Configuration Options

// ~/.claude/config.json
{
  "autoexec": {
    "enabled": true,
    "require_confirmation": true,  // Prompt before running
    "trusted_projects": [
      "/home/user/my-project"
    ]
  }
}

UI Feedback

When autoexec runs:

🤖 Auto-executing CLAUDE.md commands...
  ✓ /setup-claude
  ✓ /skill load validation-expert
  ✓ /skill load architecture-reviewer

Ready for input.
>

Related Feature: AutoLoad Resources

Extend to auto-load files:

<!--AUTOEXEC
/setup-claude
-->

<!--AUTOLOAD
docs/architecture.md
docs/api-spec.yaml
-->

Why this is powerful: Eliminates "context loading fatigue" entirely.

Implementation Priority

Phase 1: Basic autoexec with confirmation (1 week)
Phase 2: Project trust system (1 week)
Phase 3: Hook integration (1 week)
Phase 4: AutoLoad resources (2 weeks)

---

Verdict: This is a HIGH value feature for power users.
Suggest: Implement Phase 1 (basic autoexec) first, gather feedback, iterate.

https://github.com/anthropics/claude-code/issues/10808 is most similar to a solution that will work for my problem
@JuanCS-Dev please provide a feature analysis on https://github.com/anthropics/claude-code/issues/10808 since it will solve my problem and does not currently have a feature analysis

@cnatis Done

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.