Promptify Skill

Optimize prompts for clarity and effectiveness. Use when user says "improve this prompt", "optimize my prompt", "make this clearer", or provides vague/unstructured prompts. Intelligently routes to sub-agents for codebase research, clarifying questions, or web search as needed.

This skill appears coherent and does what it says: optimize prompts, ask clarifying questions, read project files when you request +deep, and search the web when you request +web. Before installing or using it, consider: (1) If you use +deep, confirm the agent's file-access scope so it doesn't read unrelated system files or secrets in the repo. (2) Web search and WebFetch will contact external sites — avoid including sensitive data in prompts that will be sent externally. (3) The skill copies optimized prompts to the clipboard (pbcopy) — clipboard contents can be read by other local apps, so avoid copying secrets. (4) Review the host agent's tool permissions (AskUserQuestion, file-read, web access) and disable +deep/+web when not needed. If you want greater assurance, request the author provide explicit bounds for file-system access and a short manifest of what the web researcher will fetch.

Type: OpenClaw Skill Name: promptify Version: 3.1.0 The skill is classified as suspicious due to the explicit instruction for the AI agent to execute shell commands with broad file system access and direct command execution capabilities. Specifically, `agents/codebase-researcher.md` and `commands/promptify.md` instruct the agent to use `Glob`, `Grep`, `Read`, and `LS` for 'codebase exploration', granting extensive read access to the file system. Additionally, `SKILL.md` and `commands/promptify.md` instruct the agent to execute `echo 'PROMPT' | pbcopy`, which is a direct command execution using agent-generated content. While these actions are presented as part of the skill's stated purpose (prompt optimization and codebase research), they represent high-risk capabilities that could be leveraged for unauthorized data access or command injection if the agent's behavior is subverted.