Lean prose that assumes Claude's competence — it never explains what JSX, base64, or a regex are — and every section (procedure, decode helper, edge cases, 'Do not') earns its place in the context budget.

Provides an authoritative detection regex, an executable copy-paste decode() function, a concrete node -e inspection tip, exact file paths, and a precise reverse-line-order Edit instruction — all copy-paste ready.

A clear 7-step sequence includes an explicit Verify step (re-read, confirm no stray markers, run typecheck/lint/build) with a fix-what-you-broke feedback loop, appropriate for destructive batch edits across multiple slide files.

No bundle files exist (references/, scripts/, assets/ all absent) and none are needed; the body is cleanly sectioned (Marker format, Procedure, helper, Edge cases, Do not) with no nested reference chains, satisfying the well-organized single-file criterion.

Names multiple concrete actions — 'Apply pending @slide-comment markers', 'perform the described edits', and 'delete the markers' — within a clearly scoped domain, matching the 'lists multiple specific concrete actions' anchor.

Explicitly answers both what the skill does (apply markers, edit, delete) and when to use it via an explicit 'Use when the user asks to...' clause with concrete triggers.

Provides natural quoted phrasings a user would actually say ('apply comments', 'process slide comments', 'apply the inspector comments') plus a concrete file-path trigger, giving strong coverage of likely variations.

Targets a narrow niche (open-slide inspector @slide-comment markers tied to slides/<id>/index.tsx) with distinct triggers unlikely to fire for unrelated skills.