v1.1.0 — manual-only default

PromptCoach no longer analyzes every prompt automatically.

• Manual invocation is now the default. The UserPromptSubmit auto-invocation hook is disabled (hooks/hooks.json ships with empty hooks: {}); coaching runs only when you invoke /prompt-analyzer:analyze-prompt.
• The original hook is preserved in the file's _disabled_UserPromptSubmit block; set auto_observe: true and restore it to coach every prompt again.
• Docs (README, CLAUDE.md, config help) updated to the manual-only default; auto_observe example set to false.
• Added CLAUDE.md with contributor guidance.

Install:

/plugin marketplace add rayapatiranjithai/prompt-analyzer
/plugin install prompt-analyzer@rayapatiranjithai-plugins

v1.0.0

First release of prompt-analyzer — a Claude Code plugin that scores and coaches how you write prompts against documented prompt-engineering standards.

Highlights

• analyze-prompt — live per-prompt coaching: a plain-language standards verdict, a standards checklist, stage-aware scoring, one focus habit at a time, cause→consequence→fix explanations, a standard-compliant rewrite, and a running Analysis Window of session stats.
• prompt-report — read-only end-of-session scorecard: best/worst prompt, improvement sparkline, focus habit, and recommendations.
• Auto-invocation hook (bypass any prompt with *), privacy modes (raw / redact_pii / hashed / metadata_only), token/time efficiency, and voice-prompt detection.

Install

/plugin marketplace add rayapatiranjithai/prompt-analyzer
/plugin install prompt-analyzer@rayapatiranjithai-plugins