engineering / skill
claude-reflect
A Claude Code plugin or mod for reviewing corrections and saving approved rules to instruction files.
At a glance
- Purpose
- A Claude Code plugin or mod for reviewing corrections and saving approved rules to instruction files.
- Runs in
- Local agent with shell
- Operating systems
- macOS, Linux, Windows
- Needs
- Runtime: Python 3.6+ for the plugin; Needs internet: Not stated; Needs accounts or API keys beyond your agent: Not stated
- Files and data
- Reads files: Yes; Writes files: Yes; Sends data to other services: Not stated
- Tool type
- Skill
- Delivery
- From GitHub
- Version
- Latest on the main branch
- Access
- Code on GitHub
- Compatibility
- According to the README: Claude Code, with the plugin on macOS, Linux and Windows (native Python, no WSL required). The mod needs Claude Code 2.1.286 or newer with mods enabled. The README says approved learnings can also go to an AGENTS.md file, which it lists as working with Codex, Cursor, Aider, Jules, Zed and Factory.
- External costs
- The repository is MIT-licensed and states no pricing. The README says each prompt the mod checks is one small model call on your own Claude account. Your coding agent or model provider may charge for usage.
- Permissions
- The plugin and mod act through Claude Code with the access you give them. According to the README, plugin hooks run at session start, on prompts, before compaction and after git commits, and commands read local session history. The README says approved learnings can update ~/.claude/CLAUDE.md across projects, project and subdirectory CLAUDE.md files, .claude/commands files, an existing AGENTS.md and locally linked guidance documents. The README says skill discovery creates command files and lists /skip-reflect to discard all queued learnings. For the mod, the README describes a model call on your Claude account and a Save control that writes to the global or project CLAUDE.md. The README suggests editing cleanupPeriodDays in ~/.claude/settings.json and gives an upgrade procedure that deletes two plugin-cache directories under ~/.claude/plugins/.
- Usage terms
- MIT License, as stated in the repository. See its LICENSE file for the exact terms.
Code on GitHub
BayramAnnakov/claude-reflect
MITA self-learning system for Claude Code that captures corrections, positive feedback, and preferences — then syncs them to CLAUDE.md and AGENTS.md.
Topics: claude-code claude-skills memory productivity self-learning
Public GitHub data fetched · 0 days old
View on GitHub ↗What it does
claude-reflect is a Claude Code plugin or mod for saving corrections and preferences to instruction files. The README says to choose one form: the Python plugin queues corrections through hooks for later processing with /reflect, while reflect-mod checks short prompts with a model and presents Save, Edit and Skip controls.The README says approved plugin learnings can be written to global, project and subdirectory CLAUDE.md files, .claude/commands files, an existing AGENTS.md, and documents reachable through local includes or Markdown links. The README lists /reflect --scan-history for past sessions, /reflect --dedupe for similar entries and /reflect-skills for recurring tasks.
The README says the mod's Save control adds a bullet to ~/.claude/CLAUDE.md or ./CLAUDE.md. The README lists /reflect-queue and /reflect-pause, and says remember: followed by a rule skips the model check. The README warns that installing both forms captures corrections twice.
Who it is for
Claude Code users who want the corrections and preferences they give Claude saved into their CLAUDE.md files.Example tasks
Run /reflect to go through queued corrections and apply, edit or skip each one.Run /reflect --scan-history to scan past sessions for corrections you made before installing.
Run /reflect-skills to find repeating tasks that could become reusable commands.
Run /reflect --dedupe to find and consolidate similar entries in CLAUDE.md.
Run /reflect --targets to see which files will be updated.