Content
78%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The content is a strong, lean overview that concentrates on FigJam-specific gotchas and operational rules Claude would not otherwise know, with excellent progressive disclosure into 13 real, well-scoped reference files. The main gaps are the absence of an explicit post-mutation verification/feedback loop and a couple of minor trims and inline code examples that would make the guidance fully copy-paste ready.
Suggestions
Add a verification step to the core workflow, e.g. "After batch or destructive mutations, call get_screenshot on the affected section to confirm the result before proceeding" — its absence caps workflow clarity.
Include one short copy-paste use_figma code snippet for the canonical text-mutation recipe so the most common operation is fully executable from this file.
Trim the round-trip justification under Deferred Tools and the skillNames logging caveat to a single sentence each.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious operational knowledge (console.log not returned, createPage throws in FigJam, upload_assets is the only image path, get_metadata unsupported) and assumes Claude's competence throughout. A few asides could be trimmed — e.g. "Six sequential ToolSearch calls is six round trips before any work happens. One batched call is one round trip." and restating the skillNames logging caveat — which places it at anchor 4 rather than 5's "every token earns its place." | 4 / 5 |
Actionability | Guidance is concrete and executable: exact tool names, the exact ToolSearch query string ("select:use_figma,get_figjam,get_screenshot,get_metadata,create_new_file"), exact error messages to expect, and precise load-then-mutate recipe pointers. Anchor 5 is not fully met because the text-mutation recipe and image upload are described at the rule level without a copy-paste code example in this file (delegated to the sibling skill's references), leaving minor gaps. | 4 / 5 |
Workflow Clarity | The sequence is clear and well-ordered: inspect with get_figjam first to obtain IDs, batch-load references and deferred tool schemas, then mutate and return IDs — with error-avoidance checkpoints built in (invalid nodeId errors, unloaded-font errors). It falls short of anchor 5 because there is no explicit verification step (e.g. "confirm with get_screenshot after mutations") and no validate-fix-retry feedback loop, which matters given the batch-modify reference. | 4 / 5 |
Progressive Disclosure | The body is a lean overview with a well-organized Reference Docs section listing all 13 references, each with a one-line scope description; every referenced file exists in ./references/ and references are one level deep. Detailed content (per-node-type APIs, colors, planning) is correctly pushed to reference files, and cross-links to the sibling skill's references are clearly signaled. Matches anchor 5: clear overview with well-signaled one-level-deep references and easy navigation. | 5 / 5 |
Total | 17 / 20 Passed |