Content
82%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.
This is a dense, well-disciplined router skill: token-efficient tables route every marketing job to a named skill with an explicit fetch mechanism, and the Rules section packs real constraints (approval gates, credential handling, output locations) with no padding. The main weaknesses are abstract 'First steps' that reference unlisted actions and a few dangling references (focused-reads.md, typed knowledge action) whose source is never stated. Concrete setup/validation detail would close the gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and telegraphic — routing tables, terse commands ('gooseworks fetch <slug>', 'Save a skill's files in /tmp/gooseworks-scripts/<slug>/'), and rules where every clause encodes a real constraint. It never explains concepts Claude already knows and every token earns its place, matching anchor 5 rather than anchor 4's 'minor instances of over-explanation'. | 5 / 5 |
Actionability | There is mostly executable guidance: concrete CLI commands with flags ('gooseworks login --device --no-wait', 'gooseworks call <provider> <path> --query'), tool names (catalog_search, catalog_fetch), and an exact save path. It stops short of anchor 5 because core steps reference things not defined in this file — 'the listed setup/status action', 'the listed typed knowledge action', 'follow focused-reads.md' — leaving gaps in the common setup path. | 4 / 5 |
Workflow Clarity | A clear sequence exists (numbered 'First steps', intent→skill→how routing tables, setup→route→fetch→playbook flow) with most checkpoints present: approval gates ('state the call count and total credits; wait for the customer's yes'), stop conditions ('Stop creation if unavailable'), and error recovery ('If a command says you are not signed in, run gooseworks login…'). It misses anchor 5 because the First steps are abstract — 'Discover the listed setup/status action and read brand/task state' has no explicit validation of what 'done' looks like. | 4 / 5 |
Progressive Disclosure | With no bundle files, the body appropriately stays an overview: routing tables plus one-level-deep, clearly signaled fetch pointers ('Fetch it with catalog_fetch, or gooseworks fetch goose-graphics in a terminal'), organized for discovery. It falls short of anchor 5 because a few references dangle without location or context — 'follow focused-reads.md' (a file inside a fetched skill, effectively a second hop) and 'the listed typed knowledge action' never say where they come from. | 4 / 5 |
Total | 17 / 20 Passed |