Content
92%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.
An exceptionally tight, well-engineered instruction skill: every section carries executable direction, the multi-round workflow is sequenced with acceptance gates and fallback paths, and the method catalog is kept out of context behind a purpose-built lookup script. The only noticeable gap is the absence of a single worked example (a brief, a return, a report section) to anchor the formats it specifies so precisely.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is telegraphic instruction with zero padding: concepts like "expert panel, stakeholder lens, Delphi" and cold readers are referenced without explanation, and every sentence directs an action or a format (e.g. "Save each return as `round-<r>-push-<n>-<method>.md`... the open push as `open`"). It assumes Claude's competence throughout; not 4 because there is no over-explanation anywhere to trim. | 5 / 5 |
Actionability | Guidance is highly executable: exact uv-run command pipelines, a fully enumerated five-part brief spec, a precise return shape ("the form as the body, then a `## Findings` list... at most five findings"), exact file-naming conventions, and an enumerated user-choice list. It falls just short of 5 because no worked example of a brief, return, or report section is provided, and what a finding's "location" looks like is left unspecified — the anchor-5 criteria ask for specific examples covering the common cases; it is well above 3 since nothing is pseudocode or vague. | 4 / 5 |
Workflow Clarity | The workflow is explicitly sequenced (Setup → Target/Start → Brief → Round → Merge → Present and Choose → Done) with checkpoints throughout: "Stop to ask first only when there is no clear target", "Nothing in the target changes until the user accepts", error-recovery loops ("Script not found means BMad is not set up here: offer the bmad skill's setup... then run it again"; the without-subagents fallback), and a per-change applied/rejected record under each round. This matches the anchor for clear sequence with explicit validation and feedback loops; not 4 because no checkpoint category is missing. | 5 / 5 |
Progressive Disclosure | The bulk content — the 9KB method catalog in assets/methods.csv — is deliberately kept out of context and served in lean slices through the verified bundle script scripts/pick_methods.py (`list`, `show`, `random`, with `--all` gated as deliberate), and the SKILL.md body references only that script plus ecosystem setup scripts, all one level deep. Clear overview, appropriately split content, easy navigation; not 4 because the split and the serving commands are both explicit and verified against the actual bundle. | 5 / 5 |
Total | 19 / 20 Passed |