Skills human-paced-web-ops

Use human-paced browser interaction patterns for web navigation and search tasks with variable delays, hover-before-click, and light randomness. Improves robustness and reduces brittle bot-like behavior while respecting website rules.

install

source · Clone the upstream repo

git clone https://github.com/openclaw/skills

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/openclaw/skills "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/1477009639zw-blip/human-paced-web-ops" ~/.claude/skills/clawdbot-skills-human-paced-web-ops && rm -rf "$T"

manifest: skills/1477009639zw-blip/human-paced-web-ops/SKILL.md

source content

Human-Paced Web Ops

Use this skill when the task involves:

Web search and browsing in dynamic pages
Multi-step page navigation that easily breaks with rigid scripts
Long-running read-and-collect workflows

Interaction Pattern

Apply these defaults unless the task needs exact deterministic clicking:

Before actions, wait for visible/interactive state first.
Use small randomized delays between actions (for example 300-1200ms).
Prefer
```
hover
```
before
```
click
```
on menus/buttons when possible.
Use small random click offset inside the same target element (for example 2-8px), not random page clicking.
Add occasional small scroll steps during long pages.
Avoid repeated fixed-interval requests; pace actions with jitter.
Every 5-10 interactions, re-check page state and URL before continuing.

Guardrails

If blocked by CAPTCHA, login challenge, paywall, or anti-bot page, pause and report the blocker URL plus required manual step.
Keep identity and session settings stable and traceable for repeatable runs.
Respect robots/terms and prefer official APIs when available.

Output Requirement

For web collection tasks, include:

What was visited (titles + links)
What was extracted
What could not be accessed (and why)
Next recoverable step