openclaw-reflect

Self-improvement layer with evaluation separation, rollback, and tiered operator gates. Observes outcomes across sessions, detects recurring patterns, propos...

What to consider before installing: - Data exfiltration: If you set ANTHROPIC_API_KEY/OPENAI_API_KEY or point to an Ollama host, the skill will send an excerpt of MEMORY.md and sample input summaries to those services for evaluation. Do not enable remote evaluators if your MEMORY.md or tool inputs contain secrets or sensitive data. - Automatic file changes: The skill can auto-append to MEMORY.md and CLAUDE.md when confidence thresholds are reached. Review the apply/rollback logic, test in a sandbox workspace, and consider requiring operator approval (avoid --auto) until you trust thresholds and outputs. - Payment capability: The repository includes a voluntary payment flow (AGENT-PAYMENTS.md) that shows how an autonomous agent could POST to a local x402 endpoint to make contributions. Ensure your agent cannot make payments without explicit human authorization and that localhost endpoints are protected. - Audit artifacts: After installation, monitor .reflect/ (outcomes.jsonl, proposals.json, pending.json, applied.jsonl, snapshots/) and review pending proposals before approving. Use rollback.js to revert undesired changes and verify snapshots are created as expected. - Prompt content: The evaluator prompts enforce exact output formats; this is by design but increases the impact of a compromised evaluator backend. If you must run evaluation remotely, prefer a private Ollama instance or rule-based fallback. If you want to proceed: install and run the skill in a restricted test workspace first, do not provide remote API keys initially, and manually review any queued proposals before approval. If you decline automatic application, run apply.js only with explicit --id --approve commands.

Type: OpenClaw Skill Name: openclaw-reflect Version: 1.0.2 The `openclaw-reflect` skill is designed for agent self-improvement, observing tool outcomes, detecting patterns, and proposing changes to `MEMORY.md`, `CLAUDE.md`, and `SOUL.md`. Its core functionality involves reading/writing to the workspace (`.reflect/`, `MEMORY.md`, `CLAUDE.md`), making network calls to external LLM evaluators (Anthropic, OpenAI, Ollama), and executing internal Node.js scripts (`scripts/*.js`). All these actions are explicitly aligned with its stated purpose, which includes safety mechanisms like evaluation separation, tiered approval, and rollback. The `SKILL.md` and `README.md` contain instructions for the agent to surface proposals and for operators to run commands, which are standard interaction patterns for OpenClaw skills. A voluntary payment request via a local x402 API is present in `AGENT-PAYMENTS.md`, but it is explicitly optional and uses a local endpoint, not an external malicious one. There is no evidence of intentional harmful behavior such as data exfiltration to unauthorized endpoints, backdoors, or malicious prompt injection attempts against the agent.