Conciseness

The body is information-dense and largely free of concepts-Claude-already-knows filler, but it runs ~560 lines for what is styled as an overview file, and sections repeat themselves: the L1–L4 image channels are fully described, then re-stated in 'Decision (run per slide)', then again across the anti-patterns; the Live preview section spends many lines on host-environment caveats. This is 'mostly efficient but could be tightened' (anchor 3) rather than anchor 4's 'minor instances of over-explanation' — the trimming opportunity is a section, not a sentence.

Actionability

Guidance is copy-paste ready throughout: nine exact CLI invocations ('uv run scripts/dump_text.py input.pptx --notes'), complete download code patterns (urllib Request with User-Agent header, BytesIO variant, curl equivalent), a grep-able QA text-integrity check, and concrete tables (typography sizes, slide dimensions, channel budget per deck type). Specific commands cover the common cases, matching anchor 5; the only near-gap is the actor({...}) subagent snippet, which mirrors harness tool syntax rather than being runnable code, but it is illustrative by design.

Workflow Clarity

The decision matrix routes every situation to a path and a 'read first' file, the pipeline order is explicit ('read → plan slide-by-slide → edit/create → validate → visual QA'), and the QA checklist supplies validation checkpoints with an explicit feedback loop ('If any of these fail, fix and re-run — don't paper over') for the batch/destructive XML operations (explode/prune/assemble). This matches anchor 5 — clear sequence, explicit validation, feedback loops, and a checklist — and clearly exceeds anchor 4's 'minor validation gaps'.

Progressive Disclosure

Structure and signaling are strong: a decision matrix with 'Read first' pointers, and a closing 'Where each detail lives' map — all one level deep. However, the three primary reference files it points to (create.md, edit.md, read.md) do not exist in the bundle (only SKILL.md and scripts/ are present), so the main disclosure links dead-end; the scripts referenced (dump_text.py, preview.ts, soffice_bridge.py, etc.) all exist. Additionally, ~130 lines of image-sourcing detail and the typography/scene-default tables are inline here even though the skill's own map says such recipes live in create.md. That is 'good structure... minor organization gaps' (anchor 4) rather than anchor 3, whose hallmarks (buried, unclearly signaled references) are absent — but the broken reference targets keep it below anchor 5's 'easy navigation'.

Specificity

The description lists many concrete actions — 'authoring slide decks', 'extracting text or structure from an existing .pptx', 'filling a .pptx template with values', 'converting a deck to PDF or images', 'splitting or merging decks', 'inspecting slides, layouts, masters, tables, images, charts, speaker notes, or comments' — comprehensive coverage of the toolkit's capabilities with no vague filler. This matches the anchor-5 example ('Extract text and tables... fill forms, merge documents, convert pages to images') and exceeds anchor 4, which permits minor coverage gaps.

Completeness

Both 'what' (produced, opened, transformed, or read; the enumerated task list) and 'when' ('Use this skill whenever...', 'Trigger on words like...') are explicitly stated, plus concrete negative triggers ('Do NOT trigger when the primary deliverable is a Word document, spreadsheet, PDF report, HTML site, or Google Slides API call'). This is the anchor-5 pattern verbatim in structure; a 'Use when' clause is present, so the completeness cap of 3 does not apply, and the exclusion list makes the 'when' more explicit than the anchor-5 example itself.

Trigger Term Quality

It includes natural user phrasings ('deck', 'slides', 'presentation', 'pitch deck', 'keynote', 'executive readouts'), the domain name ('Microsoft PowerPoint'), and the file extension ('any filename ending in .pptx'). Coverage including synonyms and extensions matches anchor 5 ('PDF files, PDFs, forms, document extraction, .pdf'); anchor 4 would require noticeably missing natural terms, which is not the case.

Distinctiveness Conflict Risk

The niche is tightly bounded to Microsoft PowerPoint (.pptx) files and it explicitly disambiguates against the nearest competing skills (Word documents, spreadsheets, PDF reports, HTML sites, Google Slides). Clear niche with distinct triggers and minimal conflict risk is exactly the anchor-5 case; anchor 4 ('minor overlap risk with closely related skills') is surpassed because the overlaps are affirmatively excluded.