Conciseness

At ~66KB / 1,260 lines for a single-URL capture workflow, the body is noticeably verbose: the ob.read() empty-vs-Error trap is repeated in Step 0, Step 2.5a, Step 4, and Rule 19; the cp1252/UTF-8 issue appears three times; and dated incident logs ("Seen 2026-10-02 and again 2026-10-09", "observed 2026-09-17 → 09-19") are embedded inline rather than in a deprecated/old-patterns section. Much of the content is genuinely non-obvious operational knowledge, so this is not anchor 1, but the padding and repetition exceed anchor 3's "some unnecessary explanation".

Actionability

Every step ships copy-paste-ready, executable guidance: the preflight probe one-liner, the note_text() guard helper, `fetch.py "<url>" --max-chars 0`, the GitHub tree/raw API recipe, the prune_recently_added function with exact regex, exact image filter lists (w_150, h_72), and the Substack archive API endpoint. Even failure paths have executable fallbacks, fully matching anchor 5.

Workflow Clarity

Steps 0–8 are clearly sequenced with explicit validation checkpoints and feedback loops: a mandatory preflight probe, the URL-dedup gate before fetching, `assert moc` and `assert anchor in t` guards, verify-the-write by re-reading frontmatter, retry-once on the transient xAI error, and SKIPPED-vs-FAIL semantics for the optional open step. Overwrites (create --overwrite) are guarded by re-reads, matching anchor 5's "explicit validation steps; feedback loops for error recovery".

Progressive Disclosure

The single bundle file (scripts/fetch.py) is real and properly referenced by path, and internal structure is good (numbered steps, tables, callout boxes). But the body itself is a monolith: the gotcha catalogue, the 13-folder routing table, the image-captioning rules, and the degraded-mode fallback table are all inlined in SKILL.md when they clearly belong in separate one-level-deep reference files — anchor 3 ("some structure but content that should be separate is inline"). Not 4 because the bulk of the detail is inline rather than split out.

Specificity

"Fetch a URL, summarize it, and save as a structured research note" plus "source of truth for tags, folder routing, and freshness updates" lists several concrete actions including the secondary MOC-maintenance operation. It falls short of anchor 5's comprehensive coverage because tweet/X handling, dedup, and the thesis-drift check are absent, and "summarize it" is bare.

Completeness

The "what" is clear (fetch, summarize, save a structured note into the Research/Library taxonomy), but there is no "when" — the second sentence ("Use the master Research Library MOC as the source of truth...") is operational instruction, not trigger guidance, so per the judging guideline a missing 'Use when...' clause caps this at 3. Not a 2 because the "what" is concrete rather than vague.

Trigger Term Quality

Terms like "URL", "summarize", "research note", and "Obsidian" are ones a user might naturally say, but common variations are missing — no "save this article", "clip", "tweet", "read later", or "capture". This matches anchor 3 (relevant keywords but missing variations) rather than anchor 4, where only a few natural terms would be absent.

Distinctiveness Conflict Risk

"Structured research note in the current Obsidian Research/Library taxonomy" with the master MOC as source of truth carves a clear niche with distinct artifacts. Minor overlap risk remains with closely related sibling skills (e.g. obsidian-daily-research, which also writes Library notes), matching anchor 4 rather than 5.