Content
88%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.
An exemplary repo-specific skill: nearly all content is non-inferable configuration knowledge, commands are fully executable with real paths, and both the dev loop and final validation workflows carry explicit checkpoints and failure-mode warnings. The only weaknesses are minor: some command repetition between the detailed sections and the Quick reference, a small "--" typo in the speedup bullets, and an unverifiable external skill reference.
Suggestions
Trim the Quick reference table to only scenarios not already shown verbatim above it (e.g., keep only the TRX-report and single-test rows) to cut duplicated tokens.
Fix the "-- **Single TFM**" bullet on line 64 to use a single dash so it renders as a list item like its siblings.
Add a one-line note on where the `run-tests` / `dotnet-test` plugin skill comes from (or inline the top MTP-vs-VSTest flag table), since the reference cannot be resolved inside this bundle.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Almost every line is repo-specific knowledge Claude cannot infer (single-threaded runner config, auto trait filters, unconditional coverage args, DOTNET_HOST_PATH), with no filler explaining known concepts. The Quick reference table repeats commands already shown and one bullet opens with a stray "--" (line 64), leaving minor trim opportunities that keep it at level 4 rather than the every-token-earns-its-place level 5. | 4 / 5 |
Actionability | Commands are copy-paste ready with real project paths ("src\Tasks.UnitTests\Microsoft.Build.Tasks.UnitTests.csproj"), TFM-specific exe paths, correct MTP flags (--filter-method, --report-trx), and explicit pitfalls ("Switching TFM without rebuilding silently runs stale binaries"). The dev-loop, per-project, and full-suite cases are all covered with executable examples. | 5 / 5 |
Workflow Clarity | Two clearly sequenced workflows (fast scoped dev loop; final validation pass) with explicit checkpoints and failure-mode guidance: "Always check the `skipped` count" for hidden regressions, "revert before final validation" for parallel overrides, "Restore parallelism settings" as step 1, and a warning not to cancel the ~9-minute full run. This matches the anchor with explicit validation steps, not merely "most checkpoints present". | 5 / 5 |
Progressive Disclosure | The body is well organized into purposeful sections and appropriately defers the exhaustive MTP-vs-VSTest flag reference to an external skill ("see the `run-tests` skill from the `dotnet-test` plugin") while keeping only the essential pitfalls inline. No bundle files exist to verify that pointer, and the 130-line single file slightly exceeds a lean overview, so this is good structure with minor organization gaps rather than the level-5 ideal. | 4 / 5 |
Total | 18 / 20 Passed |