Convert now finds the plugin location and copies prompt.md and ralph.sh
directly from templates/, ensuring consistency across projects and
automatic sync when running /ralph:convert again.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Major improvements inspired by coleam00/ai-agents-masterclass:
- prd.md: Auto-detect project context (package.json, pyproject.toml, etc.)
- NEW_PROJECT_MODE: ~10 questions for new projects
- FEATURE_MODE: ~5-7 questions for existing projects
- Uses AskUserQuestion for interactive UX
- Optional WebSearch for best practices
- convert.md: Extended functionality
- Generates .claude/settings.json with allow-list per tech stack
- Extracts real commands from package.json scripts
- Adds techStack info to prd.json
- Creates screenshots/ directory
- prompt.md: Browser verification with agent-browser CLI
- Compact refs (@e1, @e2) for minimal token usage
- requiresBrowserCheck flag per story
- Screenshot saving for UI verification
- ralph.sh: Screenshots directory + agent-browser check
- README.md: Complete rewrite with new features
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add /ralph:prd command for PRD generation with clarifying questions
- Add /ralph:convert command to convert PRD to executable JSON
- Include templates: ralph.sh, prompt.md, prd-template.json
- Include example PRD (prd-hello-api.json)
- Update marketplace.json with ralph plugin v1.0.0
- Update README with ralph documentation
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add explicit header example with HTML paths comment (Pas 6.1)
- Add v2.x → v3.0 migration instructions
- Compact all output messages (ultra-compact format)
- Remove EXCLUSE section from output
- Bump version to 3.2.0
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Hooks execute after Claude Code exits, so output is not visible.
Removed: setup command, install-hooks.sh, hooks.json, session-end-reminder.sh
Bump version to 2.2.0
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Plain text messages for SessionEnd and PreCompact hooks
- Use echo instead of printf
- Bump version to 2.1.4
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
SessionEnd hooks require { hooks: [...] } format, not direct type/command.
Fixed install-hooks.sh script and setup.md command.
Bump version to 2.1.3
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add setup.md command to install SessionEnd and PreCompact hooks
- Add install-hooks.sh script for automatic hook installation to ~/.claude/settings.json
- Hooks remind user to run /learn:analyze before session end or auto-compact
- Bump version to 2.1.2
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
New features:
- /learn:analyze now uses signal detection to save only valuable lessons
- Explicit commands: "ține minte", "remember", "memorează"
- Stated preferences: "întotdeauna", "always", "never"
- Repeated corrections (2+ times)
- Historical references: "cum am discutat", "as we discussed"
- New /learn:cleanup command for memory hygiene
- Scoring system for entry quality
- Conservative mode with confirmation
- Upgrade option to rewrite entries
- SessionEnd hook shows reminder on /exit, /clear, Ctrl+D
- Metadata tracking: source, confidence, reinforced, trigger
Changes:
- analyze.md: Added Pas 3.5 for signal detection
- cleanup.md: New command for reviewing/deleting entries
- hooks.json: Changed from Stop to SessionEnd
- scripts/session-end-reminder.sh: New reminder script
- Templates updated with metadata fields
- Documentation updated for v2.1.0
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Commands:
- /learn:analyze - Analyze conversation for patterns and gotchas
- /learn:import - Import from other memory files (auto-build-memory.md, JSON)
- /learn:reflect - Consolidate and clean memory
- /learn:status - Show memory statistics
Features:
- Stop hook suggests /learn:analyze after significant sessions
- Uses single memory file: .claude/rules/claude-learn-memory.md
- Structured format for patterns (solutions) and gotchas (problems)
- Import support for migrating existing memories
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
