Content
76%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 well-structured, highly actionable reference: every section carries executable examples and genuinely non-obvious operational knowledge. The main weaknesses are the absence of a retry/recovery feedback loop for failed files in a batch context and minor redundancy plus one ambiguous reference location.
Suggestions
Add a feedback loop to the 'Error recovery' section: show how to detect failed entries in the results array (e.g., an error/status field per result) and how to re-run just the failed files, such as re-invoking `xberg batch` on the failed paths or applying `--file-configs` overrides like `force_ocr` for the retry.
Trim redundancy: the `--format json` vs `--format text` default difference is stated in both 'Basic usage' and 'Common pitfalls' — keep it in the pitfalls list only, and drop the opening paragraph that restates the frontmatter description.
Disambiguate the final reference — state that `references/cli-reference.md` lives in the sibling `xberg` skill (as done for the python/nodejs/rust API references), since this bundle contains no `references/` directory.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and xberg-specific throughout (scheduler behavior, layout concurrency caps, file-configs key matching), with only minor trims: the opening paragraph restates the frontmatter description, and the json-vs-text default-format difference appears in both 'Basic usage' and 'Common pitfalls'. Not 5 because of that small duplication. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready bash commands, a complete `--file-configs` JSON object, a TOML config, and a complete Python `extract_batch` example covering the common cases (defaults, concurrency caps, overrides, image output, config files). | 5 / 5 |
Workflow Clarity | This is a batch operation, and the error-recovery section does give one explicit checkpoint ("Inspect results for partial content and surfaced errors rather than relying on the process exit code alone"), but there is no validate-fix-retry feedback loop — no guidance on re-running or skipping failed files — so the batch-operation cap holds it at 3. Not 4 because the sequence across sections (choose flags, run, verify) is implicit rather than a clear checkpointed workflow. | 3 / 5 |
Progressive Disclosure | Well-organized sections with bulk API detail correctly deferred to one-level-deep references; however, `references/cli-reference.md` is cited with no stated location and no `references/` directory exists in this bundle (the other three references are explicitly placed in the sibling `xberg` skill), a minor navigation gap that keeps it below 5. | 4 / 5 |
Total | 16 / 20 Passed |