Conciseness

Mostly dense, non-obvious operational guidance, but with real repetition: the anchor-loading list appears three times (Activation, 'Always load at activation', Recovery Step 1), the pipeline is described three times (Lifecycle Position diagram, task-type table, numbered 'Pipeline position'), the 'read fresh, not from recall' rule recurs several times, and the surf metaphor adds padding ('No wipeouts. No zombie sessions.'). Fits anchor 3 (mostly efficient, could be tightened) rather than 2, since most sections earn their tokens.

Actionability

Fully executable throughout: copy-paste commands ('entire status 2>/dev/null', 'ls .context-surfing/handoff-*.md 2>/dev/null'), a complete settings.json hook config with install-path variants, install commands, a verbatim handoff file template with an exact naming pattern, and a ready-to-use user-notification message. The one non-verbatim item (the subagent review brief) is still specified field-by-field: status, anchors read, concrete observations, alignment, and a continue/re-anchor/hand-off recommendation.

Workflow Clarity

The lifecycle is clearly sequenced (Activation → Riding the Wave → Drift Detection → Recovery Protocol → Exit Protocol → handoff → re-entry) with an explicit validation checkpoint (the pre-commit anchor check: quote verbatim from a fresh read → state pending action → verify trace-back), a capped feedback loop ('Allow at most one local re-anchor and one cold-context check for the same drift episode'), defined escalation branches, and a checklist-style handoff template. No validation step is missing, so it matches anchor 5 over anchor 4.

Progressive Disclosure

The single bundle file (scripts/handoff-checker.sh) is real, correctly referenced in the Hook Integration section with a copy-paste path, and section headers make the document navigable — but the skill is a ~420-line monolith with no references/ files at all. Content needed only at specific moments (the handoff template, hook setup JSON, interoperability with sibling skills, learning-signal details) is inlined rather than split into on-demand files, matching anchor 3 ('content that should be separate is inline') more than anchor 4.

Specificity

Names several concrete behaviors — 'Monitors context window health', 'Uses bounded external anchors', 'one cold-context review when self-recovery is uncertain', 'a clean handoff on real drift' — but 'bounded external anchors' leaves the mechanism vague and the monitoring outputs (drift signals, exit) are not enumerated. Fits anchor 4 (several specific actions, minor gaps) rather than 5, which requires comprehensive coverage.

Completeness

The 'what' is clear (monitor, anchor, cold-context review, clean handoff) and the 'when' is explicit — 'during large, long-running, multi-session, or explicitly context-sensitive work' — plus a negative boundary ('Do not auto-activate merely because a medium task has an intent frame and plan'), so the missing-'Use when' cap does not apply. Not 5 because the triggers are relative task categories rather than concrete user-sayable trigger phrases.

Trigger Term Quality

Contains natural terms users would plausibly say — 'context window', 'large', 'long-running', 'multi-session', 'context-sensitive', 'drift', 'handoff' — with good coverage overall. It misses common variations such as 'context is filling up', 'running out of context', or 'fresh session', so it sits between anchor 3 (missing variations) and anchor 4 (good coverage) → 4.

Distinctiveness Conflict Risk

Context-quality drift monitoring during execution is a distinct niche, made more so by the explicit anti-trigger for ordinary Medium tasks. Minor overlap risk remains with closely related sibling skills that also claim 'multi-session' work and 'handoff' territory (e.g., a session-context skill), matching anchor 4 rather than 5.