claude-code - 💡(How to fix) Fix Skill or setting to skip permission prompts for esoteric/low-risk commands [1 participants]

TL;DR

Implementing a broader allowlist pattern language or a "trust compound forms of already-allowed tools" mode could reduce repetitive permission prompts in Claude Code.

Guidance

• Introduce a glob or regex-based pattern matching for allowlists to cover variations of commands, reducing the need for repetitive prompts.
• Consider adding a "trust compound forms" mode that automatically approves commands based on previously allowed tools or commands.
• Discuss and decide on the scope of such a feature, whether it should be project-specific, session-based, or have an expiration mechanism to balance convenience and security.
• Evaluate the best integration point for this feature, such as within settings.json permissions or as a dedicated skill.

Example

No specific code example can be provided without further technical details, but the concept would involve enhancing the allowlist functionality with more flexible pattern matching and possibly a learning mechanism for compound commands.

Notes

The implementation should carefully consider security implications to avoid introducing vulnerabilities. User feedback and testing will be crucial in refining this feature to ensure it meets both usability and security requirements.

Recommendation

Apply a workaround by enhancing the allowlist with broader pattern matching or introducing a "trust compound forms" mode, as this approach directly addresses the user's need for reduced prompts without waiting for a potential full version upgrade.