Content
71%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 well-organized, command-dense body with executable examples, sensible dry-run validation around batch operations, and honest handling of the pre-v1.0 CLI's shifting surface. The main structural flaw is that the reference and asset bundle files are completely orphaned — SKILL.md never navigates to them, leaving duplicated command detail inline and the cookbook, troubleshooting, and persona content undiscoverable.
Suggestions
Add a short 'References' section linking each bundle file (e.g., '**Command reference**: See [references/gws-command-reference.md](references/gws-command-reference.md)', plus the cookbook, troubleshooting, and persona-profiles files) so the companion content is discoverable from SKILL.md.
Consolidate the repeated '(verify)' caveats into the single opening 'Verify before scripting' blockquote and trim the per-command 'check exact flags with...' lines, keeping the (verify) marker only where syntax is genuinely unknown.
Give each + helper that is currently shown as 'gws <service> +helper ...' (reply, forward, read, append, insert, meeting-prep) one full worked invocation, or point to the specific schema/help command output that supplies the flags.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with commands and tables and explains nothing Claude already knows, so it sits well above anchor 3. Kept from 5 by the verify-before-scripting caveat repeated well beyond the top blockquote — '(verify)' markers and 'check exact flags with: gws gmail +reply --help' appear ~15 times and could be consolidated into the opening warning. | 4 / 5 |
Actionability | Most commands are copy-paste executable ('gws gmail +send --to "team@company.com" --subject "Weekly Update"...', 'python3 scripts/gws_doctor.py'), and the few placeholders ('gws gmail +reply ...', 'gws sheets +read ...', 'gws calendar +insert ...') are explicitly justified by the pre-v1.0 verify-first caveat. Not 5: those helper commands lack even one full worked example each. | 4 / 5 |
Workflow Clarity | Workflows are clearly sequenced with validation for batch/destructive ops — 'Use --dry-run first, and --page-all to paginate', 'Preview, then archive read emails older than 30 days', plus doctor pre-flight and the audit -> review FAIL -> remediate loop, so the missing-validation cap at 3 does not apply. Not 5: no explicit error-recovery guidance (e.g., rate-limit 429s are listed as a limitation but no retry/backoff step is given for bulk runs). | 4 / 5 |
Progressive Disclosure | The body itself is well structured and all five scripts/ files are properly referenced, but none of the bundle's companion files — references/gws-command-reference.md, references/recipes-cookbook.md, references/troubleshooting.md, assets/persona-profiles.md, assets/workspace-config.json — are ever linked or mentioned, so that detail is unreachable from SKILL.md while overlapping command syntax sits inline. This matches anchor 3 ('references present but not clearly signaled') rather than 4 ('references mostly clear'). | 3 / 5 |
Total | 15 / 20 Passed |