Content
80%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.
The content is an efficient, well-organized set of behavioral rules with concrete thresholds and verbatim output strings, and it is appropriately compact for a single-file skill with no bundle dependencies. Its weakness is the absence of any validation or feedback checkpoints — nothing tells the model how to verify it is complying or what to do when a rule conflicts — leaving the workflow clarity at the anchor-3 level.
Suggestions
Add a lightweight verification step, e.g., 'Before finishing a pass, confirm the todo list reflects what was actually done' or an explicit self-check after each context-threshold action, to introduce feedback into the rule set.
Make rule 11 concrete: specify what 'write in batches' means (e.g., 'emit one section per tool call, then continue') so the overflow guidance is as executable as the context rules.
State how to resolve conflicts between rules (e.g., what to do when scope exceeds limits mid-task or the user overrides the scope-reduction proposal) so the sections behave as a coherent process rather than independent rules.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean, directive-style rule set with zero padding: every line is an operational rule or a quantified threshold, and it explains nothing Claude already knows, matching the 'every token earns its place' anchor. | 5 / 5 |
Actionability | Most rules are executable as written — exact thresholds (65%/100K, 75%/120K, 2h/15 files/350 lines) and verbatim output strings ('WARNING! High context consumption, consider using new session!') — but items like 'Write in batches, section-by-section' and 'output size risks overloading the context' lack concrete mechanics, fitting anchor 4 rather than 5. | 4 / 5 |
Workflow Clarity | Rules are grouped into labeled sections with clear condition→action pairs, but the content is a standing rule set rather than a sequenced workflow: there are no validation checkpoints, no feedback or error-recovery loops, and the sections do not form an ordered process, matching anchor 3 ('steps listed but checkpoints missing'). | 3 / 5 |
Progressive Disclosure | The body is under 50 lines, needs no external references, and is organized into clearly labeled sections (Planning, Context, Scope, Output, Communication), which per the rubric's simple-skill guideline earns anchor 5. | 5 / 5 |
Total | 17 / 20 Passed |