← ClaudeAtlas

probelisted

Resolve a consequential uncertainty with the smallest useful experiment before committing to a larger direction. Use when the user asks to test an assumption, check feasibility, run a bounded spike, or find out whether an approach works before building it. Define success, failure, limits, and cleanup before execution; distinguish observed results from inference. Runs authorized reversible experiments and returns evidence to the decision, without expanding into full implementation.
kennykankush/skillpack · ★ 0 · Code & Development · score 60
Install: claude install-skill kennykankush/skillpack
# Probe — Learn Enough to Make the Next Decision Probe turns one important unknown into an experiment whose result can change what happens next. The deliverable is evidence about that unknown. A prototype, script, measurement, or observed task can be the means. Stop when the experiment answers its question or reaches its declared limit. ## When to use - "Test this assumption before we commit" - "Can this actually work with what we have?" - "Run a small feasibility spike" - "What is the smallest experiment that would settle this?" - `decide` identifies a consequential unknown and the user authorizes investigating it. ## When not to use - Existing evidence or a direct lookup settles the question: use it; `scour` can verify external claims when needed. - The task is to understand existing code: use targeted reading or `devour`. - The user wants a known bug fixed or a decided feature built: do the work normally. - The task is a structural audit or a whole-journey stress sweep: use `bedrock` or `gauntlet`. - The answer would not affect a decision: explain why the proposed experiment adds little. ## Preflight — define the experiment before running it Read the existing context first. Briefly state the following in the conversation or the project's existing experiment record. Fill in real values appropriate to the question: ```text Decision this informs: <what we will do differently depending on the result> Assumption: <one falsifiable claim> Experiment: <smallest setup tha