Add meta-command for post-failure command improvement workflow

[benceruleanlu]

Currently, when a command doesn't perform as expected due to
edge cases (e.g., checking if on a feature branch but not the
correct feature branch), there's no streamlined workflow to
capture these learnings and improve the command immediately.

Example scenario: The PR-submit-local-as-new-pr.md command
initially only checked if the user was on a feature branch, but
didn't validate if it was semantically appropriate for the
changes being made. This led to PRs including unrelated commits.

Proposed solution: Create a new command like
/development:improve-command-after-failure that:

1. Captures the failure context and resolution
2. Updates the command with the fix OR
3. Documents which alternative command should have been used
4. Optionally creates a "lessons learned" entry

This would enable rapid iteration on commands based on
real-world usage patterns and edge cases.

(yes my github pat 403ed for issue creation, i'm fixing it)

[christian-byrne]

That's really smart, great idea. Anything that lets us iterate on the CC tooling over time is money.

[christian-byrne]

By the way, whenever I am using CC to help adjust any CC tooling, I start with this command

comfy-claude-prompt-library/.claude/commands/agents/AGENT-learn-claude-code-best-practices.md

Line 4 in 736336d

You are a technical educator specializing in AI-assisted development workflows. Your task is to comprehensively study Claude Code's features, tools, and best practices to help the user become more effective at using Claude Code.

to give good context