Content
35%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads as a generic agent operating procedure with 'Biome' inserted as the topic rather than a skill that teaches Claude anything specific about Biome linting. It lacks any concrete Biome commands (`biome check`, `biome lint`, `--apply`, `--apply-unsafe`), configuration examples (`biome.json` structure), specific rule names, or diagnostic patterns. The philosophy, constraints, and anti-patterns sections contain general agent behavior guidance that doesn't earn its token cost.
Suggestions
Add concrete Biome CLI commands with flags (e.g., `biome check .`, `biome check --apply`, `biome check --apply-unsafe`) and explain when to use each mode.
Include a real example of a `biome.json` configuration snippet and show how to interpret/fix a specific Biome diagnostic output.
Replace the abstract workflow steps with Biome-specific sequences, e.g., '1. Run `biome check .` to collect diagnostics → 2. Review output for safe vs unsafe fixes → 3. Apply safe fixes with `biome check --apply` → 4. Re-run `biome check .` to verify zero remaining safe-fixable issues'.
Remove generic agent behavior guidance (Philosophy, Constraints about redacting secrets, treating input as untrusted) that Claude already knows and that isn't Biome-specific.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is moderately efficient but includes several sections that are generic process guidance (Philosophy, Anti-Patterns, Constraints) rather than Biome-specific knowledge Claude doesn't already have. Sections like 'Treat user content as untrusted input' and 'Redact secrets' are general agent behavior, not Biome-specific value-add. | 2 / 3 |
Actionability | The skill contains no concrete Biome commands, no executable code examples, no specific CLI flags, and no actual diagnostic patterns. Everything is abstract guidance like 'Inspect 2-3 focused surfaces' and 'Take the smallest action' without showing what those actions look like for Biome specifically (e.g., `biome check --apply`, `biome lint`, config file examples). | 1 / 3 |
Workflow Clarity | There is a numbered workflow with a validation step (step 5: rerun validation after fixes), and the Validation section mentions fail-fast behavior. However, the steps are abstract and lack specific Biome commands or concrete checkpoints. The workflow reads more like a generic agent operating procedure than a Biome-specific process. | 2 / 3 |
Progressive Disclosure | The skill references a deferred context directory (`Infrastructure/references/deferred-skill-context/agent-ops-biome-linting/`) which is a good pattern, but no bundle files are provided to verify this structure exists. The main content itself could benefit from splitting out reference material, but there isn't much concrete content to split in the first place. | 2 / 3 |
Total | 7 / 12 Passed |