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, disciplined skill body: executable commands with verifiable expected outputs, a validated six-step workflow with explicit error-recovery loops, and clean progressive disclosure into verified one-level references. The only weakness is a small amount of dated versioning housekeeping in the install section that could be trimmed or pushed to a reference.
Suggestions
Trim the mfapy versioning discrepancy note ("Its README labels the latest change '064', but its installed distribution still reports `0.6.3'") to a single actionable sentence, or move it to references/inference.md; keep only the instruction to record the Git commit alongside the package version.
Replace the dated sentence "The pinned commit matched upstream `master` on 2026-09-30" with a pointer to the pinned revision in assets/requirements.txt; the specific date will go stale and adds no action for the user.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and directive with no explanation of concepts Claude already knows; statements like "Stable metabolite abundance does not establish isotopic steady state" are domain-specific discipline, not padding. However, the dated housekeeping paragraph ("The pinned commit matched upstream `master` on 2026-09-30. Its README labels the latest change '064', but its installed distribution still reports `0.6.3`") is time-sensitive trivia that could be trimmed, so it is efficient-with-minor-trimmables (level 4) rather than every-token-earns-its-place (level 5). | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands throughout: "uv venv --python 3.12 .venv-mfa", the install command with a concrete requirements path, and the three CLI subcommands (check, simulate, fit) with complete flags and file names. Worked examples even state expected numeric results for verification ("the analytical 95% interval for `straight` is about 67.55–72.45", the published MDV values), covering the common cases with no gaps, matching the top anchor. | 5 / 5 |
Workflow Clarity | The six-step workflow has an explicit validation step (step 2's check command "checks atom counts and conservation, fragments, tracer sums, uncertainty matrices, bounds, and steady-state mass-balance feasibility" with its limits stated), explicit feedback loops ("If a profile finds a better solution than the baseline, rerun the fit; do not publish the stale intervals", "Refine the grid if they are too coarse", "repeat with a different seed and more starts"), and a delivery checklist in step 6. This matches the anchor requiring explicit validation, error-recovery loops, and checklists rather than the minor-gaps level 4. | 5 / 5 |
Progressive Disclosure | The body is a clear overview that keeps detailed material out of SKILL.md and signals one-level-deep references exactly at the point of need: "Read [references/input-contract.md](references/input-contract.md) when preparing inputs. Read [references/inference.md](references/inference.md) before interpreting an actual fit." Both reference files exist and contain real content with no nested pointers (only external documentation links), the engine pin lives in assets/requirements.txt, and every asset/script path in the worked examples resolves to a real bundled file. Split, signaling, and navigation all match the top anchor rather than the minor-organization-gaps level 4. | 5 / 5 |
Total | 19 / 20 Passed |