Content
71%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 strong decision-support body: purpose, triggers, anti-triggers, a concrete five-step script workflow, a runnable example, and unusually valuable domain heuristics (retention differential, source-vs-influence attribution, consistent overhead allocation). Main weaknesses are the missing bundle files (the workflow is non-executable as shipped), the unspecified markdown-template-to-JSON input bridge, and inlined reference material with some duplicated sections.
Suggestions
Ship the referenced bundle files: SKILL.md references three scripts, three references, and one asset, but none exist in the bundle — as shipped, every workflow step and the References section is a dangling pointer.
Specify how the filled assets/channel_data_template.md becomes channel.json / roi.json / mix.json (format, fields, conversion step) so the script invocations in Steps 2-4 are reproducible.
Move the forcing-question library and the "Distinct from" section into a reference file (e.g., references/forcing_questions.md) and deduplicate them against "Do not use for" to cut SKILL.md length and repetition.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with domain-specific, non-obvious data ("MDF < 5:1", "A 5-point retention gap moves LTV by 30-50%", "$200k channel manager managing $4M of partner ARR is $50... per $1k ARR") rather than concepts Claude already knows. Minor trimmable redundancy keeps it at anchor 4: "Distinct from" re-covers the same six skills as "Do not use for", and the human-decides point is stated three times ("It does not pick the strategy — humans do", "The skill recommends; the human commits", "never an automatic resource reallocation"). | 4 / 5 |
Actionability | Commands are concrete and specific ("scripts/cost_to_serve_calculator.py --input channel.json --output markdown", "--profile {saas,api,enterprise-software,marketplace,hardware}", a runnable Quick example with --sample). Anchor 4 rather than 5 because of a minor gap: Step 1 fills a markdown template (assets/channel_data_template.md) while the scripts consume channel.json / roi.json / mix.json, and how to bridge the template to those input files is never specified. | 4 / 5 |
Workflow Clarity | A clear five-step sequence names the script, inputs, and outputs at each step, with built-in validation ("Flags double-counting and surfaces hidden costs", "The script validates that overhead % is consistent across channels") and sequencing discipline ("Lock 1-3 before opening 4-7"). Not anchor 5 because there is no explicit feedback loop — nothing says what to do when the calculator flags double-counting or inconsistent allocation before running the next script. | 4 / 5 |
Progressive Disclosure | The body itself is well-sectioned and its references are clearly signaled one level deep ("references/channel_economics_canon.md — Skok, Bessemer State of the Cloud, Tunguz..."), but the bundle contains none of the seven referenced files — no scripts/, references/, or assets/ directories exist — so the promised navigation targets are absent as shipped. Additionally, reference-grade material (canon citations, the ~30-line forcing-question library, and the "Distinct from" section duplicating "Do not use for") is inlined in the ~130-line SKILL.md. This matches anchor 3 ("could be better organized; content that should be separate is inline") better than anchor 4. | 3 / 5 |
Total | 15 / 20 Passed |