Content
77%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.
The content is highly actionable — complete, executable Nix derivations with concrete commands and a validated five-step workflow including error-recovery loops. Its weaknesses are repetition of the meta block, shebang substitution, and hash instructions across sections, and a monolithic structure with no reference files that pushes the SKILL.md toward 590 lines.
Suggestions
Consolidate the triplicated meta block and quadruplicated shebang-substitution snippet: state each once in a single section (e.g. metadata_requirements and common_patterns) and reference it from the templates, cutting significant token overhead.
Move the troubleshooting and common_patterns sections into references/ files (e.g. references/troubleshooting.md, references/patterns.md) with clearly signaled links from the main body, slimming SKILL.md to an overview plus quick_start and workflow.
Merge the duplicated hash-fetching instructions (quick_start pre-built block and step_2_fetch_hashes) into the single workflow step, keeping only the fetchBunDeps iteration detail there.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly code templates and commands rather than concept explanations, but there is noticeable duplication: the meta block appears three times (both quick_start derivations plus metadata_requirements), shebang substitution four times (quick_start, common_patterns, anti_patterns, step 4), and hash-fetching instructions twice (quick_start and step_2). This matches anchor 3 — mostly efficient but could be tightened by consolidating repeated patterns; not 2 because no section is padded with concepts Claude already knows, and not 4 because the repetition is more than minor trimming. | 3 / 5 |
Actionability | Fully executable, copy-paste-ready material: two complete derivation templates, concrete commands (nix-prefetch-url, nix hash convert, nix build, ldd, nix fmt, nix flake check), and specific special-case snippets (WASM assets, multiple executables, autoPatchelf for native deps). This matches the anchor-5 example of code covering the common cases, with no pseudocode anywhere; not 4 because there are no missing key details — even the fetchBunDeps hash-iteration loop is spelled out. | 5 / 5 |
Workflow Clarity | A clear five-step sequence (identify package type → fetch hashes → create files → special cases → test build) with an explicit validation step (step_5: build, run --version/--help, ldd, nix flake check), a build checklist, and feedback loops in the troubleshooting section (hash mismatch: copy the 'got:' hash and rebuild; bun.lock mismatch: fakeHash → build → update). This matches anchor 5's clear sequence with explicit validation and error-recovery loops; not 4 because checkpoints are present throughout rather than with minor gaps. The operation is a build, not destructive/batch, so no 3-cap applies. | 5 / 5 |
Progressive Disclosure | There are no bundle files (references/, scripts/, assets/ are absent), so everything — two full derivation templates, troubleshooting, common patterns, metadata requirements — is inlined in a ~590-line SKILL.md. Section headers are clear and navigation is adequate, which keeps it above anchor 2 (unstructured inline content), but content that would naturally live in separate reference files (troubleshooting, pattern library) is inlined, matching anchor 3. Not 4 because the split into a slimmer overview plus one-level-deep reference files is an obvious, unmade improvement. | 3 / 5 |
Total | 16 / 20 Passed |