Feature Request: SessionStart hooks should support automatic skill invocation and file reading

Status Closed — duplicate
Maintainer reply None cached
Activity 3 comments · opened Dec 2, 2025 · closed Dec 6, 2025

Summary

Add support for hooks to automatically invoke skills and read files at session start, allowing Claude to load project-specific context without user intervention.

Use Case

In our project (Aviado health analytics platform), we have:

  1. AI_ASSISTANT_INSTRUCTIONS.md - Critical workflow instructions, database credentials, deployment commands
  2. openai-api skill - Current OpenAI model availability (GPT-5 released Aug 2025, API parameter changes)

Currently, every session requires the user to ask Claude to:

  1. Read the instructions file
  2. Invoke the skill

This is error-prone - if forgotten, Claude may use outdated model information or miss critical project patterns.

Current Workaround (Insufficient)

{
  "hooks": {
    "SessionStart": [{
      "matcher": "startup",
      "hooks": [{
        "type": "command",
        "command": "echo 'Remember to read AI_ASSISTANT_INSTRUCTIONS.md and invoke openai-api skill'"
      }]
    }]
  }
}

This echoes a reminder, but Claude doesn't automatically act on it - the user still has to explicitly ask.

Proposed Solution

Add new hook types for SessionStart:

Option 1: type: "skill"

{
  "hooks": {
    "SessionStart": [{
      "matcher": "startup",
      "hooks": [
        {
          "type": "skill",
          "skill": "openai-api"
        }
      ]
    }]
  }
}

Option 2: type: "readFile"

{
  "hooks": {
    "SessionStart": [{
      "matcher": "startup", 
      "hooks": [
        {
          "type": "readFile",
          "path": "AI_ASSISTANT_INSTRUCTIONS.md",
          "injectAs": "context"
        }
      ]
    }]
  }
}

Option 3: type: "autoExecute" (combined)

{
  "hooks": {
    "SessionStart": [{
      "matcher": "startup",
      "hooks": [
        {
          "type": "autoExecute",
          "actions": [
            { "tool": "Read", "args": { "file_path": "AI_ASSISTANT_INSTRUCTIONS.md" } },
            { "tool": "Skill", "args": { "skill": "openai-api" } }
          ]
        }
      ]
    }]
  }
}

Benefits

  1. Consistency - Every session starts with correct context loaded
  2. Reduced errors - No risk of Claude using outdated information
  3. Better DX - Users don't have to remember setup steps
  4. Project portability - New team members get correct context automatically

Environment

  • Claude Code version: Latest (December 2025)
  • OS: macOS
  • Project type: Full-stack TypeScript/Python with Supabase Edge Functions

Additional Context

This feature would be especially valuable for projects that:

  • Use rapidly evolving APIs (OpenAI models change frequently)
  • Have complex deployment workflows
  • Require database credentials or connection details
  • Have team-specific coding standards

Thank you for considering this enhancement!

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗