Content
71%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 strong operational skill body: concrete commands, a precise flag surface, real tool gotchas, validation checkpoints, and a graceful offline fallback make it highly actionable. The weaknesses are token efficiency (redundant 12-command CLI block duplicating the flags table) and no use of bundle files to split reference material out of SKILL.md.
Suggestions
Collapse the twelve near-identical CLI examples into one canonical invocation plus the existing "All flags" table (or 2–3 example combinations), saving ~50 lines.
Move the Citations list, Dependencies table, and MSA benchmark table into a references/ file (e.g., references/tools.md) linked from the body, and ship the referenced phylogenetics_builder.py / examples/demo_tree.nwk inside the bundle.
Add an explicit error-recovery loop after tree inference (e.g., "if result.json workflow_state != completed, read contract_alerts, fix input/flags, re-run") to reach the top workflow-clarity anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with tool-specific knowledge Claude cannot be assumed to know (prank's ".best.fas" suffix, stripping "+F" for RAxML-NG, UFBoot ≥ 95 threshold) and avoids basic-concept padding. However, the CLI Reference repeats twelve near-identical full invocations (~50 lines) that duplicate the "All flags" table directly below it — a sizable tightening opportunity, not a minor trim, so it sits at anchor 3 rather than 4. | 3 / 5 |
Actionability | Fully executable throughout: exact engine commands ("iqtree2 -s aln.fa -m MODEL --prefix prefix [-bb 1000]"), a complete flag table with defaults, a demo-mode command, the ETE3 midpoint snippet, and concrete log-parsing targets like "Best-fit model according to BIC:". Copy-paste-ready guidance covers the common cases. | 5 / 5 |
Workflow Clarity | The 8-step workflow is clearly numbered with per-step skip conditions ("--aligned", "--no-trim", "--model"), upfront input validation (≥3 sequences, equal-length check for aligned input), RAxML-NG "--check" pre-validation, a "workflow_state == completed" checkpoint, and a mandatory demo fallback. It falls short of anchor 5 only because explicit validate→fix→retry error-recovery loops are not spelled out. | 4 / 5 |
Progressive Disclosure | Sections are clearly headed and navigable, but everything is inline in a single ~300-line file — the 12-entry Citations list, the dependency table, and the MSA benchmark table are reference material that could live in separate files. Body-referenced files ("phylogenetics_builder.py", "examples/demo_tree.nwk") are not present in the bundle's reference directories either. | 3 / 5 |
Total | 15 / 20 Passed |