Detect recurring tool-outcome failures and generate memory-augmentation proposals.
npx skills add https://github.com/velamints2/clawbot-lab --skill openclaw-reflect
Please help me install this Agent Skill. Skill: openclaw-reflect Source: https://github.com/velamints2/clawbot-lab/tree/main/skills/openclaw-reflect Command: npx skills add https://github.com/velamints2/clawbot-lab --skill openclaw-reflect
This Skill enables autonomous observation of tool outcomes across sessions, detects recurring failure patterns, and proposes memory-driven improvements that guide safer agent behavior over time.
End a session to trigger the full pipeline automatically.
High-intent search queries and answers about installing and using this skill.
To stop recurring tool-outcome failures, you need cross-session failure pattern detection that identifies repeated issues and proposes memory-driven improvements to guide safer agent behavior over time. This skill observes outcomes and classifies recurring patterns automatically.
Evaluator-backed self-improvement works by validating proposed memory updates against an evaluator before applying them. It auto-applies low-risk changes to memory files while queuing higher-risk proposals for operator review, ensuring safe agent behavior modification.
The best way to safely apply memory updates is using tiered validation that applies low-risk changes to MEMORY.md and CLAUDE.md automatically, while routing high-risk proposals to SOUL.md for operator approval. Pre-change snapshots provide rollback support.
Operator gating is optional and only required for high-risk changes queued to SOUL.md. Low-risk memory augmentations are auto-applied to MEMORY.md and CLAUDE.md without operator intervention, balancing autonomy with safety.
Yes, you can roll back memory augmentation changes using pre-change snapshots stored in the .reflect directory. The rollback support feature captures state before applying updates, allowing you to restore previous configurations if performance degrades.
Cross-session recurrence checks are needed when agents experience repeated tool-outcome failures across multiple sessions. This pattern detection identifies recurring issues that require memory-driven improvements to prevent future failures and increase reliability.
CAPHTECHGenerate an observation checklist covering five baseline failure-mode checks.
NOMARJReview Sigil quarantined items and approve or reject them via CLI.
bugrabilgeDesign benchmark test suites for evaluating LLM agent reliability.
Shamrock2245Extract experiences into reusable semantic patterns for Shamrock Trading Bot's memory system.
