Openclaw Super Skills

Decomposes complex user requests into executable subtasks, identifies required capabilities, searches for existing skills at skills.sh, and creates new skill...

This skill is conceptually reasonable (decomposing tasks and locating or creating skills), but it has several practical risks you should consider before installing or enabling it: - The SKILL.md assumes the Skills CLI (npx skills) and Node/npx are available and instructs global installs (npx skills add -g -y). Those commands will download and run third‑party code—review and vet any package before installing it on your system. - The manifest lists no required binaries or environment variables, but the workflow expects and may ask for sensitive credentials (email credentials, Slack webhooks/tokens, OAuth keys). Confirm how credentials will be provided, stored, and protected; prefer short‑lived tokens and avoid entering permanent secrets without validation. - Because the skill can create and install other skills automatically, restrict its ability to perform global installs or to run without your explicit approval. Consider running it in a sandbox or with autonomous invocation disabled until you verify behavior. - Verify the skill's source and provenance (no homepage provided and source is 'unknown'). If you plan to use it, require that any discovered skills come from trusted owners on skills.sh and review code before installation. If you want to proceed: ensure node/npx are present, run the agent in a restricted environment, disable automatic installs, and audit any skills the agent proposes to add. If you prefer lower risk, decline installation until the author supplies a trustworthy source, explicit runtime requirements, and a clear credential-handling policy.

Type: OpenClaw Skill Name: openclaw-super-skills Version: 1.0.0 The task-decomposer skill bundle provides instructions for the agent to automatically search for and install third-party skills using the 'npx skills add' command with the '-y' flag, which bypasses user confirmation. While this functionality is aligned with the stated goal of automating workflow setup, it creates a significant security risk by allowing the agent to execute global installations of unverified code from the skills.sh ecosystem based on its own search results. This pattern (found in SKILL.md and README.md) is highly susceptible to prompt injection and supply chain attacks, though no explicit evidence of intentional malice or hardcoded malicious payloads was found.