using-superpowers

The skill instructs users that Skill tool invocation is mandatory and non-negotiable before any response, including clarifying questions. This override behavior is not explicitly warned to end users, creating a risk that users may unknowingly surrender normal judgment and decision-making authority to a system-level policy they didn't explicitly consent to.

fix Add a prominent user-facing warning at the beginning of the skill documentation that clearly states: (1) This skill modifies interaction behavior to mandate skill invocation, (2) Users should understand they are agreeing to this policy, and (3) Users retain the right to override this by explicitly instructing otherwise.

The phrase '1% chance a skill might apply' is an extremely low and subjective threshold for mandatory action. This vagueness could lead to compulsive over-invocation of skills for nearly any user request, reducing utility and creating decision fatigue.

fix Replace the vague '1% chance' threshold with specific, measurable criteria. Define clear categories (e.g., 'if the task involves code generation, data analysis, or architecture design') that indicate when skill invocation is genuinely required versus optional.

The repeated use of absolute language ('absolutely must,' 'not negotiable,' 'not optional,' 'cannot rationalize your way out') creates psychological pressure that removes user agency. This coercive framing could cause users to abandon legitimate judgment or override their own expertise.

fix Reframe the policy in permissive language that preserves user choice: 'strongly recommend invoking relevant skills' instead of mandatory language. Clarify that users can choose to skip skill invocation if they have documented reasons, and include examples of legitimate exceptions.

The skill creates a system-level policy that effectively mandates behavior based on a low, subjective threshold ('1% chance'). This overrides legitimate user judgment and expertise, especially problematic when users have domain knowledge that makes skill invocation unnecessary or counterproductive.

fix Establish clear exceptions and override mechanisms: (1) Document when skill invocation should be skipped (e.g., for expert users, time-sensitive contexts, or when user explicitly declines), (2) Add a mechanism for users to set skill invocation preferences, (3) Make the policy advisory rather than mandatory.

The skill begins by establishing that 'user instructions always take precedence' (line 20-26), but the earlier sections establish mandatory skill invocation that contradicts this. This narrative inconsistency creates confusion about whose instructions actually take precedence, potentially enabling the skill to override user intent.

fix Resolve the contradiction by clearly stating: (1) User explicit instructions take precedence over skill policies, (2) Provide concrete examples of how conflicts are resolved (e.g., 'if user says skip the skill, the skill is skipped'), (3) Ensure the skill invocation mandate is truly subordinate to user control.

The 'Red Flags' table attempts to categorize normal reasoning patterns (gathering context, exploring incrementally, remembering lessons) as 'rationalizations' to be suppressed. This could inhibit legitimate critical thinking and cause users to override their own expertise without proper deliberation.

fix Rephrase the red flags section to guide rather than suppress reasoning: (1) Replace 'STOP—you're rationalizing' with 'consider whether a skill applies,' (2) Acknowledge that context-gathering and incremental exploration are valid in certain scenarios, (3) Add nuance: 'Skills are useful for X, but not required for Y.'