Mostly efficient but padded with background Claude already knows (e.g. defining what a leap second is, restating the 86400-seconds-per-day POSIX definition, and explaining the 2012 NTP kernel-hang incident); the body could be tightened without losing actionable content.

Provides fully executable, copy-paste-ready Python (time.monotonic() duration fix, freezegun harness) plus concrete fixes per anti-pattern row and a worked end-to-end example, not pseudocode.

The 'How to use this reference' and 'Worked example' sections give a clearly sequenced audit process (decide relevance, locate bug class, apply fix, assert against the testable-behaviours table, simulate), with explicit validation via the assertion step and a named caveat about what freezegun cannot verify.

SKILL.md is a concise overview that signals a single one-level-deep reference (references/smear-strategies-and-history.md) that exists and contains the deeper history/smear detail, with clear navigation; the reference file itself does not nest further.

Lists multiple concrete actions/capabilities: cataloging the 23:59:60 UTC insertion, the Google/AWS leap-smear alternative, and four named bug classes (time_t non-monotonicity, negative durations, NTP cascading, cross-node clock skew) each with a monotonic-clock fix and a freezegun simulation.

Clearly answers both 'what does this do' (pure-reference catalog of leap-second mechanics and the bugs they cause) and 'when should Claude use it' via an explicit 'Use when auditing time-sensitive code...' clause, plus a redirect to dst-transition-reference for the DST case.

Covers natural terms users would say when they need this skill: 'auditing time-sensitive code', 'financial timestamping', 'distributed logs', and 'NTP-driven schedulers', plus the explicit second-by-second progress framing.

Occupies a clear niche (leap-second mechanics and bugs) with distinct triggers, and explicitly disambiguates from the more common daylight-saving-transition skill, making a wrong-skill trigger unlikely.