Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is an exceptionally well-structured, concise, actionable workflow with strong validation checkpoints. However, its three reference files are missing from the bundle and it embeds an author-specific absolute workflow path, breaking the progressive-disclosure contract.
Suggestions
Ship the referenced bundle files (reference/subagent-prompt.md, reference/sources.md, reference/report.md) or remove the dead links — currently every sub-agent prompt, sources list, and report template the body relies on is unreachable.
Replace the hardcoded /Users/mi/claude-workspace/.mimocode/workflows/deep-research-pro.js absolute path with a bundle-relative reference (e.g. scripts/deep-research-pro.js) so the scripted-workflow alternative works in any environment.
Verify each link in the body resolves to a real file before publishing; dangling references force the model to improvise the very templates the skill promises to lock down.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — no explaining what a sub-agent or WebSearch is, tight budget tables, and imperative one-liners like 'Spawn one sub-agent per angle in a single message'. Every token earns its place. | 3 / 3 |
Actionability | Gives concrete executable guidance: a specific `date +%Y-%m-%d` Bash command, a depth-mode table with hard budgets, a directory layout, a verbatim copy-paste `workflow({...})` invocation block, and explicit per-phase file contracts. | 3 / 3 |
Workflow Clarity | A clearly sequenced Step 0 → Phase 1–5 pipeline with explicit validation checkpoints: Phase 4 gap-check against brief.md, 'If a sub-agent fails...note it and move on', a critique pass for deep mode, and resume-from-checkpoint handling — exactly the validate/feedback-loop pattern the rubric rewards. | 3 / 3 |
Progressive Disclosure | The body signals one-level-deep references — [reference/subagent-prompt.md](reference/subagent-prompt.md), [reference/sources.md](reference/sources.md), [reference/report.md](reference/report.md) — but NONE of these files exist in the bundle (no references/, scripts/, or assets/ directory), and it also points to a machine-local absolute path (/Users/mi/...deep-research-pro.js). Per the rubric, progressive_disclosure must be scored against the actual bundle structure; dangling, non-existent references are the worst case. | 1 / 3 |
Total | 10 / 12 Passed |