Lean and focused throughout with no padding or explanation of concepts Claude already knows; every section (tool guidance, matching fields, disambiguation, examples, guardrails) earns its tokens.

Names the exact tool (`experiment-list`), enumerates the specific fields to scan (name, status, feature_flag_key, recency), and prescribes the follow-up `experiment-get` call — concrete and copy-ready despite being instruction-only.

Clear sequence from search to match handling with explicit one/multiple/no-match branches and a validation checkpoint ("Confirm with the user by name before destructive actions") plus a "When NOT to search" guardrail.

A simple, single-purpose skill under 50 lines with no need for external references; it is well-organized into clearly signaled sections, qualifying for the top anchor under the simple-skills note.

Lists multiple concrete actions — resolving natural-language references to concrete experiment IDs, browsing `experiment-list`, and disambiguating multiple matches — all in third-person voice, matching the top anchor.

Explicitly answers both what (resolve references to experiment IDs) and when via a detailed "Use when..." clause enumerating named, loose, relative, and status-filtered references, satisfying the top anchor.

Rich coverage of natural user phrasing including quoted examples ("split test demo", "the signup experiment", "my pricing test") plus relative references ("latest", "most recent") and status filters, matching the good-coverage anchor.

Clear experiment-resolution niche with an explicit carve-out ("not feature-flag tools") and distinct triggers, making misfires for other skills unlikely.