Auto-approve DeepWork skills in Claude Code settings

[Copilot]

DeepWork-generated skills required manual permission approval in Claude Code on each invocation. This adds automatic permission grants to .claude/settings.json during sync.

Changes

Core implementation:

• Added sync_skill_permissions() method to AgentAdapter base class (default no-op)
• Implemented in ClaudeAdapter to append Skill(name:*) entries to permissions.allow
• Integrated into sync workflow: collects skill names during generation, syncs after hooks

Behavior:

• Preserves existing permissions
• Idempotent (prevents duplicates)
• Platform-specific (Claude only; Gemini has no skill permission system)

Example

After running deepwork install --platform claude, .claude/settings.json now includes:

{
  "permissions": {
    "allow": [
      "Skill(deepwork_jobs:*)",
      "Skill(deepwork_jobs.define:*)",
      "Skill(deepwork_jobs.implement:*)",
      ...
    ]
  }
}

Skills are immediately usable without permission prompts.

Original prompt

Skills that deepwork creates should automatically be approved for permissions in the local .claude (claude can do this for sure)

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[nhorton]

Already merged this