Conciseness

Most of the body is genuinely novel operational guidance (drift signal catalogs, recovery/exit protocols, handoff template), but several padded or duplicated sections remain: the extended ocean-wave metaphor ("Paddling in / The peak / The shoulder / The close-out", "No wipeouts. No zombie sessions"), the pipeline enumerated twice ("Lifecycle Position" and "Pipeline position"), the pre-commit anchor check described in full twice (Monitoring Paradox and "Riding the Wave" step 2), and a closing "Principles" section restating earlier content. This is mostly efficient with unnecessary explanation to trim — anchor 3 — rather than the several padded sections of anchor 2 or the minor trims of anchor 4.

Actionability

The body gives copy-paste ready commands (`entire status`, `ls .context-surfing/handoff-*.md 2>/dev/null`, `touch '<handoff-path>.consumed'`, install commands, the settings.json hook block) and a complete handoff file template, with an explicit fallback when the `context-monitor` subagent file is absent. However `learning-aggregator --deep` is invoked with no definition, path, or fallback, a concrete gap that keeps this at anchor 4 rather than the fully self-contained coverage of anchor 5.

Workflow Clarity

Every multi-step process is clearly sequenced with explicit validation and feedback loops: Activation (numbered anchor-loading steps with standalone fallback), Recovery Protocol (Step 1 re-read, Step 2 reconcile with explicit branch outcomes, bounded by "at most one local re-anchor and one cold-context check"), Exit Protocol (Step 1–3 with the full handoff checklist), and the next-session resume procedure. The pre-commit anchor check (quote verbatim from a fresh read → state pending action → verify trace-back → enter Recovery Protocol on failure) is an explicit validation gate before side-effecting actions, matching anchor 5. No destructive or batch operation cap applies.

Progressive Disclosure

The scripts/handoff-checker.sh hook is properly externalized and referenced from both the frontmatter hook registration and the Hook Integration section (verified as a real file in the bundle), with no nested references. But the ~420-line body is fully inline: the handoff template, the intent-framed-agent relationship/cadence section, and the pipeline interop details are natural candidates for references/ files that do not exist in the bundle, matching anchor 3 (content that should be separate is inline) rather than anchor 4's appropriately split content.

Specificity

"Monitors context window health", "uses bounded external anchors, one cold-context review when self-recovery is uncertain, and a clean handoff on real drift" lists several distinct actions, but they are phrased in pipeline-internal jargon rather than plain concrete operations, so it matches anchor 4 rather than the comprehensive plain-language coverage of anchor 5.

Completeness

The description explicitly answers both what ("Monitors context window health... uses bounded external anchors, one cold-context review... clean handoff") and when with concrete trigger conditions ("during large, long-running, multi-session, or explicitly context-sensitive work") plus an explicit negative activation boundary ("Do not auto-activate merely because a medium task has an intent frame and plan"). This matches anchor 5; the when-clause is explicit and specific, not the vaguer anchor 4 case.

Trigger Term Quality

"large, long-running, multi-session, or explicitly context-sensitive work" and "context window health" are natural phrases users would say, but common variations like "context window filling up", "running out of context", or "session handoff" are missing — good coverage with a few natural terms absent, matching anchor 4 rather than the comprehensive synonym coverage of anchor 5.

Distinctiveness Conflict Risk

"Context window health" during long-running, multi-session work is a clear niche, and the do-not-activate clause disambiguates from intent-framed-agent — but the description does not distinguish it from the closely related session-context skill in its own pipeline, so anchor 4 (mostly distinct, minor overlap risk with closely related skills) fits better than anchor 5.