Content
86%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.
A dense, highly actionable body that respects the reader's intelligence and covers usage, output shapes, model selection, config, programmatic access, and troubleshooting without padding. The main defect is progressive disclosure: two of the four cited reference files do not exist in the skill's bundle, leaving the 'see references/...' pointers unusable.
Suggestions
Ship references/cli-reference.md and references/advanced-features.md in the skill bundle (or remove those pointers) so the cited navigation paths actually resolve; only the python/nodejs API pointers are valid, and those are explicitly delegated to a sibling skill.
Add an explicit validation checkpoint to the workflow (e.g., after extraction, check that result.tables is non-empty and row counts look sane before tuning --layout-confidence or switching --layout-table-model).
The Node.js access section is a single sentence ('Node.js mirrors this...'); either add a minimal mirrored snippet or move the mention entirely into the reference file to avoid a half-covered surface.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean throughout: every section carries tool-specific facts (flags, output shapes, model trade-offs) with no explanation of concepts Claude already knows, and code blocks are minimal but complete. The only non-content tokens are the generated-file comment header, which is metadata rather than padding. | 5 / 5 |
Actionability | Fully executable guidance throughout: copy-paste-ready 'xberg extract' commands with concrete flags, a jq pipeline, a complete model-selection table, a working Python snippet, and symptom-to-flag remediation in the failure-modes section covering the common cases. | 5 / 5 |
Workflow Clarity | The single-action workflow (choose output format, choose table model, run, adjust confidence) is sequenced clearly as a decision path, and 'Common failure modes' provides error-recovery guidance. Not 5 because validation checkpoints are implicit — e.g., there is no explicit step to verify 'tables[]' is non-empty and sane before switching models or thresholds. | 4 / 5 |
Progressive Disclosure | The body is well-sectioned and defers bulk detail appropriately, but scored against the actual bundle: 'references/cli-reference.md' and 'references/advanced-features.md' are cited yet no references/ directory exists in the skill, so those navigation paths dangle. Not 4 because broken reference paths undermine navigation even though the signaling itself is clear; not 2 because the body is well-organized, not monolithic, and does not inline content that belongs in separate files. | 3 / 5 |
Total | 17 / 20 Passed |