Content
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
An exemplar content body: actionable, well-sequenced with validation feedback loops, and cleanly split into a lean overview plus a single well-signaled reference. The one caveat is the product-name inconsistency (description says 'OpenWork' while the body and references use 'craft-agent'/'qwen-code'), which is not penalized by these rubric dimensions but would confuse a reader.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and dense — code blocks, command lists, and bulleted guidance with no filler explaining concepts Claude already knows; every section earns its tokens. | 3 / 3 |
Actionability | Fully executable: concrete `rg` log-search patterns, `bun run` commands, and exact `mcp__chrome_devtools.*` tool-call sequences that are copy-paste ready. | 3 / 3 |
Workflow Clarity | The 7-step Harness Loop is clearly sequenced with explicit validation checkpoints ('Verify through the same path', re-run reproduction, focused typechecks) and an improvement feedback loop. | 3 / 3 |
Progressive Disclosure | A concise overview body with one clearly signaled, one-level-deep reference — 'Read `references/harness-principles.md` when...' — and the referenced file exists and matches the routing condition. | 3 / 3 |
Total | 12 / 12 Passed |