Content
81%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 tight, highly actionable procedure body: an eight-step sequence with lint-until-clean validation, explicit quality gates, and honest-state rules that constrain judgment rather than pad it. The main gaps are the unverifiable external references (templates, contract, and example all live outside this bundle) and script invocations given without full command lines.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — no space is spent explaining what workflows or boards are, and steps like "Write the flow-config block first — including unit and unit_outcome — and lint it before writing any prose" are pure directive. A few rhetorical asides could be trimmed without loss ("the first board is usually the most informative artefact the team has had in months", "not an embarrassment to paper over"), which keeps it at anchor 4 (efficient with minor trims) rather than anchor 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Guidance is concrete throughout: exact file paths ("steps/<id>.md", "platform/context/<pack>.md", "items/<slug>/item.md", "reviews/YYYY-MM-DD-scaffold.md"), exact fields ("exit evidence", "a rung with its reason", "run: human with an empty run_ref", "escalation contract for anything at rung 2+"), and named commands (flow_lint.py, flow_board.py, flow_next.py). It is not a 5 because the script invocations lack usage details (no full command line or arguments) and the templates that define the file formats live outside this bundle, so the guidance is not fully copy-paste ready on its own. | 4 / 5 |
Workflow Clarity | The eight steps are explicitly sequenced (skeleton → config linted first → step contracts → platform stubs → item cards → history → board generation → brief) with built-in validation and a feedback loop ("flow_lint.py until clean, then flow_board.py, then flow_next.py", "lint it before writing any prose"), plus a dedicated Quality Gates checklist covering lint-zero-violations, stated unit/unit_outcome, sensible first run card, and complete step contracts. This matches the top anchor: clear sequence, explicit validation steps, feedback loop, and a checklist for a complex process. | 5 / 5 |
Progressive Disclosure | The body is well organized (Outcome, Inputs, Workflow, Rules, Quality Gates, References) and the References section clearly signals one-level-deep pointers to templates, the workflow-definition contract, an example workspace, and the next skill. It is not a 5 because the referenced paths ("skills/flow-driven/templates/", "skills/flow-driven/references/workflow-definition.md", "skills/flow-driven/example/fiy-content-engine/") live outside the skill's own bundle — no references/, scripts/, or assets/ directories exist here — so none of the referenced material is verifiable, and the References section mixes file pointers with sibling-skill navigation. | 4 / 5 |
Total | 17 / 20 Passed |