Content
82%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, lean skill body with executable examples and effective anti-pattern guidance. Its main flaws are an orphaned YAML keyword block that wastes tokens, dead eval-scenario links, and a missing error-recovery loop around the tool-discovery verification step.
Suggestions
Delete the stray YAML frontmatter block after the verification command — it duplicates the SKILL.md description, serves no purpose in the body, and reads as a copy-paste mistake.
Remove or fix the three evals/scenario-*/task.md links, since the evals/ directory does not exist in the bundle; dead links break navigation and mislead the reader.
Add one or two lines of error-recovery guidance after the verification command (e.g., what to check if the tool is not listed or is named `filename_exportname` unexpectedly), turning the verify step into a true feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — a minimal quick-start snippet, a two-line package mindset table, and terse WHY-annotated anti-patterns with BAD/GOOD pairs, assuming Claude's competence throughout. It is not a 5 because of the orphaned YAML block ("description: custom tools opencode SDK plugin execute handler abort signal Zod schema...") that sits unexplained after the verification command, duplicating frontmatter content and adding pure token waste; not a 3 because that block is the only significant trimming needed. | 4 / 5 |
Actionability | Guidance is fully executable: a copy-paste-ready tool template ("import { tool } from \"@opencode-ai/plugin\"" with a complete execute handler), concrete file locations (".opencode/tool/*.ts", "~/.config/opencode/tool/*.ts"), a runnable verification command ("bun run opencode run \"list tools\""), and BAD/GOOD code pairs for each anti-pattern. Common cases are covered, matching the anchor-5 example format. | 5 / 5 |
Workflow Clarity | The Quick Start gives a clear sequenced path (create the file → restart for auto-discovery → filename=naming rule) and closes with a verification step, so most checkpoints are present. It is not a 5 because there is no error-recovery guidance — nothing says what to do if the tool is not discovered or is misnamed after running the verify command; not a 3 because validation is present, just without an explicit feedback loop. | 4 / 5 |
Progressive Disclosure | The body is a well-organized overview with one-level-deep, clearly labeled references ("advanced-patterns.md — patterns, anti-patterns, SDK examples", "sdk-api.md", "plugin-api.md", plus two example .ts files), all of which exist in the bundle. It is not a 5 because the three eval scenario links (evals/scenario-0/task.md through scenario-2) point to files that do not exist in the bundle — dead navigation links; not a 3 because the actual reference structure is appropriately split and clearly signaled. | 4 / 5 |
Total | 17 / 20 Passed |