Knowledge

OpenClaw Skill Validation Pattern

skills/openclaw-skill-validation-pattern.md


title: OpenClaw Skill Validation Pattern category: skills tags: [openclaw, skills, validation, video-analysis] summary: OpenClaw skills should avoid hard-coded model choices, load local configuration explicitly, and validate integration through non-destructive checks. created: 2026-05-25 17:18:00 CST updated: 2026-05-25 17:18:00 CST base_confidence: 0.66 lifecycle: draft lifecycle_changed: 2026-05-25 provenance: extracted: 0.76 inferred: 0.24 sources:

  • ~/.codex/sessions/2026/04/22/rollout-2026-04-22T17-50-44-019db499-63f1-7c63-9ca8-d2a25a9ad7c6.jsonl
  • ~/.codex/sessions/2026/05/13/rollout-2026-05-13T17-56-50-019e20c4-8302-75d0-bf88-9a6d7e650188.jsonl

OpenClaw Skill Validation Pattern

OpenClaw skill work in Codex history highlights a practical rule: validate the integration path and local configuration without forcing destructive service restarts unless required. ^[inferred]

Lessons

  • Skills should not hard-code LLM model choices when the host agent/tooling provides model configuration or overrides. ^[extracted]
  • Local environment values can live in machine-local config files so repo or skill logic remains portable. ^[extracted]
  • Validation should check scripts, configuration presence, backend/frontend reachability, and documented behavior. ^[extracted]
  • For PDF/vision skills, a robust pipeline may combine text extraction, page images, embedded image regions, vector layout extraction, OCR, and model vision. ^[extracted]

Related