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.
A highly actionable, executable skill body with a clear gated check-first workflow and disciplined, concrete language. Its weaknesses are structural: the reference-manager subcommand catalog is inlined where a one-level reference file would serve progressive disclosure, and there is no post-install verification step to close the feedback loop.
Suggestions
Move the ngs_reference_manager subcommand catalog (and per-subcommand output descriptions) into a references/ file (e.g., references/reference-manager.md) and link to it with a one-level, clearly signaled reference.
Add an explicit post-install validation step, e.g. 'After installing, re-run ngs_preflight.py --pipeline <name> to confirm all tools resolve' to create a validate→fix→retry loop.
Trim the exhaustive runner-type enumeration and duplicate command examples to reduce token cost without losing coverage.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is information-dense throughout — no concept explanations Claude already knows, no padded prose — with concrete flags and outputs earning their tokens. Minor trimming is possible (e.g., the exhaustive runner enumeration 'bulk RNA-seq counts/QC, scRNA FASTQ-to-count, generic DNA, germline DNA, somatic DNA, UMI panel, ATAC, ChIP/CUT&RUN, amplicon, and shotgun' and ten near-identical ngs_reference_manager invocations), which fits 'Efficient; minor instances of over-explanation that could be trimmed' rather than the lean 5 or the padded 3. | 4 / 5 |
Actionability | Every command is copy-paste executable from a stated working directory ('From the repo root'), covering the common cases: --list, --tool, --profile, --pipeline, install-plan output dirs, guarded install via --install-missing --yes, and the full reference-manager subcommand set. This matches 'Fully executable; copy-paste ready code or commands; specific examples cover the common cases'; nothing drops it to 4. | 5 / 5 |
Workflow Clarity | The 'Existence Check Order' gives a clear numbered sequence (PATH check → Python imports → package managers → indexes → emit install plan → install only on request), and the existence checks themselves are validation. Gating is explicit ('Install only when explicitly requested by the user', the shell companion 'exits without installing unless NGS_RUN_INSTALL_COMMANDS=1'). It falls short of 5 because there is no post-install verification/re-check loop and no error-recovery guidance if a check fails — 'Clear sequence with most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | The body is well-sectioned (Existence Check Order, Script, Install Strategy, Report) and flat, but no bundle files exist (references/, scripts/, assets/ are absent) and everything is inlined — notably the ten-command ngs_reference_manager catalog with per-subcommand output explanations, which is exactly the content that 'should be separate is inline'. This matches anchor 3; 4 would require the bulk of that detail to live in clearly signaled reference files. | 3 / 5 |
Total | 16 / 20 Passed |