Content
72%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 body is exemplary in token efficiency and structure — a deliberately loose skill that trusts the model and wastes nothing. Its weakness is the executable middle: the reading step (which gh commands), the analysis mechanics, and the output shape are all left unspecified, leaving genuine ambiguity about how to perform the core of the task even accounting for the stated intent to stay loose.
Suggestions
Add the concrete reading step, e.g. `gh issue list` and `gh pr list` with the relevant --state/--limit flags, so the input-gathering is executable rather than implied.
Include one small example of an output grouping (e.g. a two-line sample of "Takeable now" vs "Blocked on a decision") as a starting point that still leaves shape open.
Spell out the execution order as a short numbered sequence (read → weigh → render → write to ~/.prp/<key>/ → report choices) so the workflow doesn't have to be reconstructed from prose.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | "Incubating. Deliberately underspecified." and "Read the tracker with `gh`. Then decide what is worth showing and how to show it." are maximally lean — no padded explanations, no concepts Claude already knows, every section earns its tokens. This matches the lean-and-efficient anchor exactly. | 5 / 5 |
Actionability | Concrete anchors exist ("Read the tracker with `gh`", "write the file to this project's `~/.prp/<key>/` store and print a link", five specific considerations like "Blocked on a decision, or takeable now"), but the executable middle is missing: no gh commands, no example output, no grouping scheme. Some concrete guidance but incomplete matches anchor 3 rather than 4. | 3 / 5 |
Workflow Clarity | A sequence is discernible (read tracker → weigh considerations → decide shape → write to store and offer link → report choices) but it is implicit across sections rather than stated as steps, with no checkpoints. Anchor 3 (sequence present but implicit/checkpoints missing) fits; not 4 because the order must be inferred, and the read-and-offer operation is not destructive so no cap applies. | 3 / 5 |
Progressive Disclosure | The body is under 50 lines, needs no external references, and is well organized into clear sections (The job / What this maintainer cares about / Shape / Report your choices), meeting the simple-skill exception for a 5. The conditional "If a `helm-canvas` skill is available, follow it" is a clearly signaled pointer, not a buried reference. | 5 / 5 |
Total | 16 / 20 Passed |