Content
81%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-engineered skill body: executable tool calls with payloads, concrete numeric design rules, an explicit pre-order QC gate, and an unusually good troubleshooting feedback loop for the reverse-assembly path. Main weaknesses are the un-invoked QC script in Step 3, minor rule duplication between the design steps and the gotchas checklist, and no reference layer for reference-type material (enzyme sites, high-fidelity overhang sets).
Suggestions
Add a one-line invocation example for the QC gate in Step 3, e.g. `python scripts/cloning_qc.py --golden-gate --enzyme BsaI --parts PART1,PART2`, since the script's CLI is otherwise undocumented in the body.
Trim the Step 4 gotchas that verbatim restate Step 1/2 design rules, keeping only genuinely new items (fragment generation still needs primers).
Move the enzyme recognition-site table and high-fidelity overhang-set details into a short `references/` file and link it, to get the body closer to an overview-plus-pointers structure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and dense throughout — numeric rules ('Overlap length 15–40 bp', 'Overlap Tm ≈ 48–65 °C'), tool outputs, and enzyme sites with no conceptual padding about what cloning is. Not a 5 because 'Step 4 — Gotchas' re-states rules already given in Steps 1–2 (internal Type IIS sites, non-unique overlaps, repeats/secondary structure, Tm imbalance), a redundancy that could be trimmed. | 4 / 5 |
Actionability | Three fully executable `tu run` commands with realistic JSON payloads and expected return fields, plus concrete decision guidance — this is near copy-paste ready. Not a 5 because Step 3 cites `scripts/cloning_qc.py` without any invocation example; the CLI syntax exists only in the script's docstring, so the QC step is not executable as written in the body. | 4 / 5 |
Workflow Clarity | Clear Step 0–4 sequence (pick method → design → QC → gotchas checklist) with an explicit validation gate ('QC before ordering', PASS/WARN screening) and a genuine feedback loop with error diagnosis: 'If the assembly reports that the fragments do not chain, digest the inputs individually with DNA_virtual_digest... Getting 1 means the enzyme name or circular is wrong — not that the plasmid lacks sites.' This matches the level-5 anchor (explicit validation, error-recovery loop, checklist); the level-4 anchor's 'minor validation gaps' does not apply. | 5 / 5 |
Progressive Disclosure | Well-sectioned body with a real bundle script correctly placed at `scripts/cloning_qc.py` (verified present and matching the described checks), and hand-offs to sibling skills. Not a 5 because everything else is inlined — the enzyme-site table, overhang-set guidance, and QC details have no one-level-deep reference files, so structure is good but not split for progressive loading; a 3 would require buried references or misplaced bulk content, which is not the case. | 4 / 5 |
Total | 17 / 20 Passed |