Content
42%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.
The body is well-organized and grounded in named Conductor artifacts, but it is a ~380-line monolithic conceptual document: generic philosophy and benefits padding, abstract rather than executable instructions, and no progressive disclosure despite abundant material (artifact specs, checklists, directory layout) that belongs in referenced files. It reads more like internal project documentation than an operational skill. Trimming the conceptual sections and splitting reference material into bundled files would improve every dimension.
Suggestions
Cut the Core Philosophy, Benefits, and Institutional Memory sections (~70 lines of generic rationale Claude can infer) and keep only the operative principles in a short list.
Move the per-artifact Contents/Update-when reference and the Context Validation Checklist into files under references/ (e.g. references/artifacts.md, references/validation-checklist.md) linked one level deep from SKILL.md.
Make the workflow phases executable: give the concrete steps and commands for each phase (Context, Spec, Plan, Implement) and add explicit verify-then-proceed checkpoints, including what to do when validation flags stale or conflicting artifacts.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Quotes: "Context-Driven Development treats project context as a first-class artifact managed alongside code...", "New team members onboard faster with explicit context", "Decisions and rationale are preserved / Context survives team changes". The Core Philosophy and four Benefits sections (~50 lines) explain generic principles and obvious consequences that add little instructional value, and the Anti-Patterns Problem/Solution format pads simple one-line advice. Not a 1 because the artifact sections do carry project-specific information Claude would not know; not a 3 because the Benefits and Philosophy sections in particular are clearly trimmable padding across a ~380-line body. | 2 / 5 |
Actionability | Quotes: "Run `/conductor:setup` to create all artifacts interactively", "Move feature from 'planned' to 'implemented' in product.md", "1. Check if existing dependencies solve the need / 2. Document the rationale...". Named commands, named files, and numbered procedures give some concrete guidance, but much of the direction stays abstract ("Ensure changes in one artifact reflect in related documents", "Configure your IDE to display context files prominently") with no examples, templates, or exact commands. Not a 4 because key operations — updating an artifact, running setup on a brownfield repo — lack executable detail; not a 2 because the named files and /conductor:setup steps are genuinely actionable. | 3 / 5 |
Workflow Clarity | Quotes: "1. Context Phase... 2. Specification Phase... 3. Planning Phase... 4. Implementation Phase", "Before starting any track: 1. Read all context artifacts / 2. Flag any outdated information", and the Context Validation Checklist with "[ ] product.md reflects current product vision". The phases and a pre-implementation checklist exist, but each phase has no concrete steps or commands beyond /conductor:setup, and there are no error-recovery or feedback checkpoints (what to do when validation fails, when artifacts conflict). Not a 4 because the checkpoints are implicit/abstract rather than explicit verify-then-proceed gates like the anchor's "**Validate**: ... 4. If errors: fix and re-validate". | 3 / 5 |
Progressive Disclosure | The body is a single ~380-line document with no bundle files (no references/, scripts/, or assets/ exist) and no external references at all. Section headers are clear and consistently organized (## Artifact Relationships, ## Directory Structure, ## Context Validation Checklist), which lifts it above a wall of text, but long reference material — per-artifact content lists, the validation checklist, session-continuity procedures — is inlined in SKILL.md where the rubric expects it split into one-level-deep referenced files. Not a 2 because the in-section structure is good and nothing is buried; not a 4 because essentially all detail lives in the top-level file with zero disclosure layers. | 3 / 5 |
Total | 11 / 20 Passed |