Skill Distiller (Reference)

Complete skill compression documentation — all options, modes, and calibration details (~2,500 tokens).

This is an instruction-only reference for compressing skills and is broadly coherent with that purpose, but it references running 'ollama list' and checking GEMINI_API_KEY / OPENAI_API_KEY even though the skill metadata lists no required binaries or environment variables. Before installing or using this skill: (1) confirm how your agent will execute the provider-detection steps — will it run local commands like 'ollama list'? (2) don't expose API keys unless you understand and trust where they'll be used; ask the maintainer to add required env vars/binaries to the metadata so the scope is explicit; (3) test with dry-run/--dry-run and with non-sensitive sample skills first; and (4) if your platform enforces env-var access controls, ensure those protections are in place. If you need higher assurance, request the author clarify how LLM calls are performed given disable-model-invocation:true and add declared requirements to the registry entry.

Type: OpenClaw Skill Name: neon-skill-distiller-reference Version: 0.2.1 The skill bundle is a comprehensive reference for 'Skill Distiller,' a tool designed to help an AI agent compress verbose skill definitions to save context window space. It outlines logic for section classification, token-level importance scoring using information theory, and the use of mathematical symbols (MetaGlyph) for brevity. While it instructs the agent to check for environment variables (GEMINI_API_KEY, OPENAI_API_KEY) and write calibration logs to a local directory (.learnings/skill-distiller/), these actions are consistent with its stated purpose of LLM-based processing and self-improvement, with no evidence of malicious intent or data exfiltration.