Content
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A lean, well-organized overview with excellent progressive disclosure and verified references, but the body defers execution detail to references and only names its validation gate without a recovery loop.
Suggestions
Add a short worked example or inline the key procedure (e.g. one paper flowing through extraction to a promoted claim) so the body is executable without immediately jumping to references.
Spell out evidence-gate failure handling as an explicit validate->fix->retry loop (e.g. 'If a claim lacks the required fields, do not promote it; revise the note and re-check before any Writing handoff').
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean (~40 lines) with no concept over-explanation or padding; every section (flow, outputs, rules, read-next) earns its place and assumes Claude's competence. | 3 / 3 |
Actionability | It gives concrete output paths and a specific ruleset (e.g. promoted claims must carry 'Evidence Record ID, source type, claim strength, allowed wording, and forbidden stronger wording'), but the actual procedures for extracting claims and building the canvas are deferred to references, leaving the body incomplete for execution on its own. | 2 / 3 |
Workflow Clarity | The pipeline diagram ('Sources/Papers/ -> Knowledge/ -> Writing/ -> Maps/literature.canvas') gives a clear sequence and the 'evidence gate' is an explicit checkpoint, but the body lacks a validate->fix->retry feedback loop and step-level detail for what happens when a claim fails promotion. | 2 / 3 |
Progressive Disclosure | It is a concise overview with a well-signaled 'Read next' section listing six one-level-deep references, all of which exist in ./references/, giving easy navigation without nested lookups. | 3 / 3 |
Total | 10 / 12 Passed |