Content
92%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, operationally complete skill body: fully executable commands with explicit chemistry and search bounds, a clearly sequenced workflow with genuine validation checkpoints and error-recovery guidance, and clean one-level-deep progressive disclosure into five purpose-built reference files. The only notable weakness is mild redundancy in its limitation caveats and scattered version pins, which cost some token efficiency without hurting correctness.
Suggestions
Consolidate the repeated computational-vs-experimentally-validated caveat into a single statement in 'Validation and limits' and trim its restatements in sections 1–4 to free tokens without losing the point.
Collect version pins (primer3-py 2.3.1, Python 3.13, BLAST+ 2.17.0) in one place — e.g., a short 'Verified versions' line or in references/sources.md — and reference it from the install section, so future updates touch one location.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and skill-specific (script flags, coordinate conventions, exit-code semantics, chemistry assumptions) and assumes domain competence — no basic-concept explanations. It falls short of a 5 because the computational-vs-experimentally-validated caveat recurs across several sections ("not a blanket claim of primer quality", "they do not validate a biological assay", "this is not empirical validation", "Label untested candidates accurately") where one consolidated statement would do, and version pins (primer3-py 2.3.1, Python 3.13, BLAST+ 2.17.0) appear inline rather than collected in one place. It is above a 3 because the over-explanation is minor relative to the volume of genuinely non-obvious operational guidance. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands with every flag and concrete file paths for design, thermodynamics, and both screening engines, plus a runnable demo ("--template "$SKILL_DIR/assets/demo-template.fasta" --preset qpcr --config ..."). Common cases are covered by presets (pcr 100–1000 bp, qpcr 70–200 bp), concrete chemistry values (--mv-conc 50 --dv-conc 1.5 ...), exit-code semantics (0/1/2), and per-field output interpretation (potential_off_target, incomplete, etc.). | 5 / 5 |
Workflow Clarity | A clearly sequenced four-step workflow (design → thermodynamics → screening → validation/delivery) preceded by a request-routing table, with explicit validation checkpoints ("Inspect engine_explanations when no candidates are found", "Read the JSON scientific status even after exit 0") and feedback loops ("Change a biologically justified constraint and rerun", "resolve mapping, reference, sequence, and search problems before drawing specificity conclusions"). This matches the anchor-5 pattern of explicit validation steps with error-recovery loops. | 5 / 5 |
Progressive Disclosure | The body is a workflow overview that delegates depth to one-level-deep, well-signaled references — each verified to exist — with purpose stated at the point of reference ("Read [specificity.md] before making a specificity claim", "Use [input-contract.md] for file schemas", "Read [thermodynamics.md] before interpreting these values"). Reference files cross-link only to sibling references and external URLs, with no buried nesting; content split (workflows, thermodynamics, specificity, input schemas, sources) is appropriate to the domain complexity. The only quibble is the prose mention of the upstream `tests/primer-design/` suite, which is not part of this bundle, but it is contextual rather than a navigation pointer. | 5 / 5 |
Total | 19 / 20 Passed |