| name | claude-reflect |
| description | Self-learning capture system. Hooks detect correction patterns ("no, use X", "actually...", "use X not Y") and queue them; user runs /reflect to review and apply queued learnings to CLAUDE.md files. Use when the user mentions remembering something for future sessions, after a feature lands, or when context is about to compact with queued items. |
Claude Reflect — Self-Learning Capture
Two-stage system that helps Claude Code (and FLOSSI0ULLK harness-mates) learn from user corrections.
How It Works
Stage 1: Capture (Automatic via hooks)
Hooks detect correction patterns and queue them to ~/.claude/learnings-queue.json.
Stage 2: Process (Manual via slash command)
User runs /reflect to review and apply queued learnings to CLAUDE.md or FLOSS/docs/agent-memory/ files.
Source plugin
Full plugin (hooks, scripts, commands) is installed at:
C:\Users\kalis\.claude\plugins\cache\claude-reflect-marketplace\claude-reflect\3.1.0\
This skill projection just describes the user-facing surface; the actual capture loop runs inside Claude Code via the marketplace plugin. For cross-harness propagation (Codex, Gemini, OpenCode), upstream the capture pattern as a separate hook per harness — track as a meta-workflow follow-up.
Available Commands
| Command | Purpose |
|---|
/reflect | Process queued learnings with human review |
/reflect --scan-history | Scan past sessions for missed learnings |
/reflect --dry-run | Preview changes without applying |
/reflect-skills | Discover skill candidates from repeating patterns |
/skip-reflect | Discard all queued learnings |
/view-queue | View pending learnings without processing |
When to remind the user
- They complete a feature or meaningful work unit
- They make corrections you should remember for future sessions
- They explicitly say "remember this" or similar
- Context is about to compact and queue has items
Correction detection patterns (what the hook watches for)
High-confidence:
- Tool rejections (user stops an action with guidance)
- "no, use X" / "don't use Y"
- "actually..." / "I meant..."
- "use X not Y" / "X instead of Y"
- "remember:" (explicit marker)
Learning destinations
| Destination | Use for |
|---|
~/.claude/CLAUDE.md | Global learnings (model names, general patterns) |
./CLAUDE.md (project) | Project-specific learnings (conventions, tools, structure) |
./CLAUDE.local.md | Personal learnings (machine-specific, gitignored) |
./.claude/rules/*.md | Modular rules with optional path-scoping |
~/.claude/rules/*.md | Global modular rules |
~/.claude/projects/<project>/memory/*.md | Auto memory (low-confidence, exploratory) |
commands/*.md | Skill improvements (corrections during skill execution) |
FLOSSI0ULLK-specific destinations (additional to upstream)
Promote captured learnings to these surfaces when the learning is durable and cross-agent:
| Destination | Use for |
|---|
FLOSS/docs/agent-memory/feedback/*.md | Cross-agent feedback memories (the canonical projection materializes to .agent-surface/memory/AGENT_MEMORY.md) |
FLOSS/docs/agent-memory/project/*.md | Project-state memories that survive across sessions |
FLOSS/docs/agent-memory/user/*.md | User profile / role / preferences |
FLOSS/CLAUDE.md / workspace CLAUDE.md | Project orientation updates that need to ride at session start |
FLOSS/docs/research/<date>-<topic>.md | When the learning is research-grade and load-bearing for multi-agent work |
For substrate-level decisions captured this way, prefer routing through the consensus gateway (flossiullk-consensus MCP) and emitting a provenance packet rather than overwriting CLAUDE.md silently.
Example interaction
User: no, use gpt-5.1 not gpt-5 for reasoning tasks
[hook silently queues this to ~/.claude/learnings-queue.json]
(later, end of session)
You: I noticed a few corrections this session. Want to run /reflect to review?
User: /reflect
[command opens the queue, shows each candidate with proposed destination, asks for accept/edit/reject]
Boundary
- This is opportunistic capture — it catches things the user said in passing.
- It is NOT a replacement for explicit memory writes via
FLOSS/docs/agent-memory/.
- It is NOT a replacement for ADR updates when the correction is architectural.
- For correction tracking that needs cross-agent surface, materialize the learning into
FLOSS/docs/agent-memory/ (which the materializer projects out to all harnesses) rather than relying on a single harness's queue.
Provenance
Original skill: community/claude-reflect v3.1.0. Adapted for FLOSSI0ULLK: added the FLOSS/docs/agent-memory/* destination table + the substrate-routing boundary so correction capture doesn't bypass consensus / provenance for load-bearing changes.
