Content
76%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 highly actionable and token-efficient, with executable commands verified against the bundled scripts and a clean section structure. The main gap is the absence of validation/confirmation checkpoints around destructive send and batch label-modify operations, which caps workflow clarity.
Suggestions
Add an explicit confirmation/validation step before irreversible operations — e.g. echo the resolved recipients/subject before `send`, or require a `--confirm` flag, and verify a draft exists after `send-draft`.
Consider moving the Gmail Query Syntax and Common Label ID tables into references/ files (linked from SKILL.md) to slim the main body and push detail one level deep.
Trim redundant inline comments in code blocks (e.g. "# Mark as read (remove UNREAD label)") since the command flags are self-explanatory.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — copy-pasteable commands with minimal prose and no explanations of what email/Gmail is — but a few inline comments ("# Mark as read (remove UNREAD label)") are mildly redundant and could be trimmed. | 4 / 5 |
Actionability | Every operation is a fully executable CLI invocation with real flags that match scripts/gmail.py (verified --limit, --format, --to/--cc/--bcc, --html, --from, --add-label/--remove-label), copy-paste ready and covering common cases. | 5 / 5 |
Workflow Clarity | Although well-organized for a simple CLI skill, sending emails and bulk label modifications are destructive/irreversible operations with no validation or confirmation checkpoint (e.g. verify recipient before send, confirm draft after send-draft), so workflow_clarity is capped at 3 per the destructive-operations rule. | 3 / 5 |
Progressive Disclosure | A single well-organized SKILL.md with sections (Setup, Commands, Query Syntax, Label IDs, Token Mgmt) and real script paths; no nested references. The sizable query-syntax and label-ID tables could plausibly live in references/ but inlining is reasonable for this scope, so it sits just below a 5. | 4 / 5 |
Total | 16 / 20 Passed |