Upwork

Optimize Upwork profiles, write tailored proposals, identify client red flags, draft professional messages, and advise on pricing and client retention strate...

This skill appears coherent and focused on drafting and advising for Upwork. Before installing or using it: (1) do not paste Upwork credentials or confidential client NDA text into the agent — the skill forbids credential use but anything you paste could be included in prompts; (2) follow the human-in-the-loop rule — never auto-send proposals or messages; (3) verify the publisher/source if provenance matters to you (the registry owner is present but homepage/source is unknown); (4) watch for future updates that add install steps, env vars, or code — those would materially change the security posture; and (5) remember outputs are advisory and may be inaccurate or non-compliant with Upwork rules—always validate before acting.

Type: OpenClaw Skill Name: upwork Version: 1.0.0 The skill bundle is benign. All files are documentation (markdown) or metadata, with no executable code or external network calls. The `SKILL.md` file explicitly defines 'Critical Rules' that prohibit the AI agent from performing malicious actions such as auto-sending messages, accessing credentials, automating Upwork actions, or scraping data. These rules are consistently reinforced throughout the other documentation files (`legal.md`, `communication.md`, etc.), demonstrating a clear intent to prevent misuse and ensure human-in-the-loop operation. There is no evidence of prompt injection attempts, data exfiltration, persistence mechanisms, or obfuscation.