OpenClaw Documentation Expert — Live Fetch

Fetches live from docs.openclaw.ai on every question. 395 pages indexed, working scripts (not echo placeholders), CI-style selftest for decision-tree rot, wo...

What to consider before installing: - The skill performs live HTTP fetches only to docs.openclaw.ai (llms.txt, llms-full.txt, and <page>.md). If you need a strict offline or limited-network environment, block or review network access. - It records routing misses (timestamp | question | fetched path) to ~/.openclaw/openclaw-docs/misses.md. That file lives on the local filesystem; if you are concerned about storing copies of user queries, inspect or relocate that behavior before use. - The included scripts are short, readable Bash using curl/grep/sed; review them yourself if you want to be extra cautious. selftest.sh will make many HEAD/GET requests when run (CI-style), which can be noisy or rate-limited by the docs host. - No credentials or secrets are requested, and the skill does not contact unknown external endpoints — a good sign. If you combine this skill with other skills that do request credentials or that can access arbitrary files, re-evaluate combined risk. - Overall this appears coherent and proportional to its stated purpose; install if you trust docs.openclaw.ai and are comfortable with the local misses log and occasional verification traffic.

Type: OpenClaw Skill Name: openclaw-handbook Version: 1.4.0 The skill is a documentation assistant for OpenClaw designed to fetch live content from docs.openclaw.ai. It utilizes a set of simple bash scripts (fetch.sh, find.sh, search.sh, verify.sh) to retrieve and search markdown files via curl, and a logging script (record_miss.sh) to record routing failures locally in ~/.openclaw/openclaw-docs/misses.md for future improvement. The instructions in SKILL.md establish clear safety boundaries, such as forbidding self-modification and ensuring citations, and the overall logic is strictly aligned with the stated purpose of providing accurate documentation without any indicators of malicious intent or data exfiltration.