Content
85%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, highly actionable CLI skill with excellent error-recovery and verification workflows and almost no wasted tokens. Its main weakness is structure across files: the monolithic 295-line body should be split into reference files, and the one reference it does make is a broken link to a nonexistent checklists.md.
Suggestions
Create the referenced checklists file (and fix the path — the body links to reference/checklists.md but the conventional bundle directory is references/), so 'For multi-step operations, see...' does not point the agent at a missing file.
Move the per-resource command catalogs (Resend email, tinker/remote commands, billing and usage key schemas) into one-level-deep reference files, keeping SKILL.md as a lean overview with well-signaled links.
Deduplicate the repeated notes (redeploy-after-secret-changes appears in both 'Configuration and Resources' and 'Secrets'; filesystem ephemerality appears in both 'Build and Deploy' and 'Domains and Storage') to tighten the body toward a 5 on conciseness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and imperative, and nearly every line carries Cloud-specific facts Claude would not know (ephemeral filesystems, R2 bucket-level visibility, 15-minute build timeout, flag combos, the sodium-only extension requirement). It falls short of 5 only through minor redundancy — the redeploy-after-secret-changes rule and filesystem ephemerality each appear in two sections. | 4 / 5 |
Actionability | Fully executable throughout: copy-paste-ready commands for every workflow (ship, deploy, monitor, secrets, domains, purge, vanity-domain), exact flag combos per operation type, jq extraction examples for usage output, and runtime option discovery via `cloud <command> -h` instead of hardcoded signatures. | 5 / 5 |
Workflow Clarity | Deployment paths are matched to the task, every deploy must be followed by `cloud deploy:monitor -n`, failures route to `deployment:logs` with instructions to read the failed step's output, and the "When a Command Fails" loop (read error, check status, fix once, then stop and ask) provides explicit validation and feedback loops, including confirmation gates for destructive commands. | 5 / 5 |
Progressive Disclosure | Section structure and headers are good, but the 295-line body inlines per-resource command catalogs (Resend email, tinker, remote commands, billing key schemas) that belong in separate reference files, and the body's single outward reference — "see [reference/checklists.md](reference/checklists.md)" — points to a file that does not exist in the bundle (no references/, scripts/, or assets/ directories are present). | 3 / 5 |
Total | 17 / 20 Passed |