Using Superpowers 0.1.0

Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions

This skill enforces a strict policy: always invoke the Skill tool before responding if any skill might apply. On the positive side, it is coherent with its stated purpose and has no install or secret requirements. However, consider these issues before installing: - It forces the agent to check skills even when unnecessary, which can cause many skills to be loaded and increase data exposure and attack surface. Ask how your platform sandboxing works: when a skill is loaded, can it read user-provided context or call external endpoints? If so, this policy may accidentally leak data. - The SKILL.md forbids using the Read tool and assumes specific tools (e.g., Skill tool, TodoWrite). Confirm those tools exist and that forcing that exact workflow won't break your environment. - This instruction can degrade agent behavior (constant interruptions to check skills) and may be incompatible with platforms that expect clarifying questions first. Recommendations: only install if you understand how your platform loads and sandboxes skills and you want strict, centralized enforcement of skill usage. If you proceed, monitor which skills get loaded when this policy is used and consider restricting which skills are permitted to be auto-invoked. If you need more assurance, request an implementation that is less absolute (e.g., suggest skill checks for specific categories or make invocation user-confirmed) or provide evidence that loaded skills cannot exfiltrate data.

Type: OpenClaw Skill Name: using-superpowers-0-1-0 Version: 1.0.0 The skill bundle contains aggressive meta-instructions in SKILL.md designed to override the AI agent's default decision-making logic. It uses high-pressure language ("ABSOLUTELY MUST", "not negotiable") to force the agent to prioritize tool invocation (the 'Skill' tool) before any response or clarifying question, even if the relevance is as low as 1%. While these instructions appear intended to enforce a specific workflow, the use of prompt-injection techniques to bypass agent autonomy and the explicit instruction to avoid standard file-reading tools ('Never use the Read tool on skill files') are characteristic of behavioral manipulation.