Conciseness

The body is dense, prescriptive, and assumes competence ("No ellipsis in labels. No text over text."), with no explanations of concepts Claude already knows. Not 5 because of trimmable redundancy: the nine-kinds table is duplicated verbatim in `references/kinds.md`, the fixed hue meanings appear both inline and in `references/grammar.md`, and the intro sentence listing agents ("Claude, Cursor, Codex, OpenClaw, Hermes...") earns little.

Actionability

Guidance is highly concrete: exact commands ("Run `node templates/figure.mjs out.svg`"), exact formats ("`var(--m-rose-ink, #b0344d)`", "Dashed is `5 4`"), numeric constraints (640-unit grid, 40 KB cap), and a worked caption example. Not 5 because the central executable artifact the method depends on ("Start from `templates/figure.mjs`" — steps 6 and 7) is absent from the bundle, so the primary draw-and-run path cannot be executed as shipped.

Workflow Clarity

The 10-step Method is clearly sequenced with explicit validation and feedback loops: step 7 "It stops on a label that is too small", step 8 "Run the checks in `references/checks.md`. Fix every finding", step 9 render and inspect at 100% and 50%, plus a full "Check before you finish" checklist and a rebuild-verify loop ("build again and confirm that the committed SVG equals the script output"). This matches the top anchor with checkpoints, error-recovery loops, and a checklist.

Progressive Disclosure

The Files section clearly signals three real one-level-deep references (`references/kinds.md`, `references/grammar.md`, `references/checks.md`), each referenced at the right point in the body, and each exists with substantive content. Not 5 because `templates/figure.mjs` is listed in Files and used in the Method but is missing from the bundle, and some reference content (kinds table, hue meanings) is duplicated inline rather than split cleanly.

Specificity

"Draws static SVG figures for lessons, READMEs and manuals" names the domain and output format concretely, and enumerates specifics ("nine figure kinds, fixed hue meanings, eight arrow styles, and checks for legibility and markup"), matching the 'several specific actions; minor gaps' anchor. Not 5 because it is essentially one capability (drawing figures) with attribute detail rather than comprehensive coverage of distinct actions.

Completeness

The what is explicit ("Draws static SVG figures for lessons, READMEs and manuals in the AI Engineering from Scratch look") and the when is given as an explicit, equivalent trigger list ("Trigger phrases: 'draw a figure', ..."). This matches the anchor 'clearly and explicitly answers both what AND when with concrete trigger phrases'; score 4 would require the when to be less explicit, which it is not.

Trigger Term Quality

Trigger phrases "draw a figure", "blueprint figure", "sequence figure", "state machine figure", "figure for this lesson", "add a diagram to the manual" are natural phrasings a user would say, giving good coverage. Not 5 because common variations like a standalone "diagram"/"SVG"/".svg" mention are absent, leaving a few natural terms missing.

Distinctiveness Conflict Risk

"static SVG figures" plus branded look terms ("blueprint figure", "AI Engineering from Scratch") carve a clear niche with mostly distinct triggers. Not 5 because generic phrases like "draw a figure" and "add a diagram" could also plausibly match charting or general diagram skills, leaving minor overlap risk.