Conciseness

The body is lean and assumes Claude's competence — no concept explanations, no padding — but two facts are stated twice: invariant 9's "facts that arrive while you work wait for the next promotion" duplicates the Environment-facts bullet "facts proposed since then wait for the next promotion", and situation 3's "Saving nothing hands the same batch to every later run" duplicates the closing bullet "If you do not save, the facts stay waiting and a later promotion run gets them again". These minor redundancies could be trimmed, fitting anchor 4 rather than the every-token-earns-its-place anchor 5.

Actionability

Fully concrete, executable guidance: exact command syntax ("tessl memory pending --topic <topic> --out-dir <dir>", "tessl memory save --topic <topic> --file <note file> --note-version <version>", "--allow-empty"), exact file names ("note.md", "facts.json", ".cloud-launch/inputs.json"), a precise line format ("- <fact> (<who>, <YYYY-MM-DD>)"), quoted failure text ("Could not write the promotion input"), and how to source the topic. Per the rubric's scoring note, an instruction-only skill with guidance this specific and executable earns the top anchor.

Workflow Clarity

The workflow is a first-match decision list ("Take the first situation that applies", four ordered situations) plus explicit validation and recovery loops for every failure mode of this destructive batch operation: version-mismatch refusal → re-run pending and rewrite; over-48 KB refusal → trim per invariant 8 and re-save; folder refusal → new folder, second refusal → stop and report; no-answer save → retry, whose refusal then confirms success. The batch-operation cap at 3 does not apply because validation checkpoints and feedback loops are present throughout.

Progressive Disclosure

There are no bundle files, and none are needed: the skill is single-purpose and every section (situations, invariants, CLI behaviors, judgment rules) is load-bearing each time it runs, so splitting content out would only add a mandatory file read. The body is well-sectioned with a clear hierarchy (What done looks like → Invariants → Environment facts → Judgment), making navigation trivial — the spirit of the self-contained well-organized-skill allowance.

Specificity

The description lists multiple concrete actions covering the full promotion lifecycle: "merge restatements", "apply newer facts and retractions", "drop what can be checked live", "keep each line attributed and the note under 48 KB", and "save the note, which clears the batch". Coverage is comprehensive rather than having minor gaps, matching the anchor 5 example's breadth.

Completeness

Both halves are explicit: a detailed "what" (merge, apply, drop, attribute, size-cap, save, clear batch) and a concrete "Use when..." trigger clause naming the launching job and the input it passes. This matches the anchor 5 pattern of answering what AND when with concrete trigger phrases, and the missing-'Use when' cap is not triggered.

Trigger Term Quality

Natural domain terms are comprehensively covered with synonyms: "pending facts", "project memory topic"/"TOPIC", "memory promotion job", "run", "note", "batch", "Tessl". The trigger phrase "Use when Tessl's memory promotion job launches a run for a topic and passes it as the input TOPIC" states exactly how this skill is naturally invoked; no natural domain term is missing.

Distinctiveness Conflict Risk

The description carves out a clear niche — promoting one Tessl project-memory topic's pending facts, triggered only by that job's run and its TOPIC input — so overlap with any other skill is minimal. No other skill description would plausibly claim this territory.