Content
68%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 very lean, mostly actionable body with excellent token efficiency and a clear read-order structure. Its two real defects are missing validation checkpoints in the run-embed workflow and — more seriously — a read order that points at three reference files that are absent from the bundle, breaking progressive disclosure entirely.
Suggestions
Ship the referenced files references/eda.md, references/charts.md, and references/reports.md (or remove them from the read order) so the disclosed navigation actually resolves.
Write out the single-mode commands in full instead of 'ags.py analysis run-eda ... --type pygwalker', and add one usage example for export_plotly_html/export_pygwalker_html.
Add an explicit checkpoint after run-eda (e.g., 'Verify data/eda_hub.html exists before embedding; if missing, re-run with --type eda-hub') to close the workflow's validation gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~40-line body is lean and assumes competence throughout — commands are given as flag-annotated snippets, the read order is three numbered lines, and lines like 'Embed `data/eda_hub.html` as primary §数据 interactive surface' carry maximum information per token. No concept explanations or padding exist to trim, so it matches the anchor-5 example rather than the minor-over-explanation anchor 4. | 5 / 5 |
Actionability | Concrete executable guidance dominates: the full 'ags.py analysis run-eda --db-path DB --output-dir DIR/data --type bundle --workspace WS --hypothesis-id HID' command and a real import path are present. However, the single-mode commands use '...' elision ('run-eda ... --type pygwalker') and the Python block shows only imports with no usage, leaving minor gaps that keep it below copy-paste-ready anchor 5. | 4 / 5 |
Workflow Clarity | A sequence is inferable (run-eda at stage 2, read references in order, export charts, embed eda_hub.html at stage 5) and stages are listed, but validation checkpoints are entirely absent or implicit — there is no step to verify the bundle or eda_hub.html was actually generated before embedding. This matches 'sequence present but checkpoints missing or implicit' rather than the minor-gaps anchor 4. | 3 / 5 |
Progressive Disclosure | The body itself is well-sectioned with a clearly signaled one-level-deep read order, but all three referenced files — references/eda.md, references/charts.md, references/reports.md — do not exist in the bundle (no references/, scripts/, or assets/ directories are present), so the disclosure chain is broken at runtime. Dangling references fail navigation outright, which is below the 'references present but not clearly signaled' anchor 3. | 2 / 5 |
Total | 14 / 20 Passed |