Content
88%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 configuration guide: exact commands, config snippets, per-client paths, and a verify-then-troubleshoot loop with real validation steps. The only deductions are minor redundancy in auth guidance and catalog detail inlined in the body rather than split into a reference file.
Suggestions
Remove one of the two statements discouraging query-string credentials — the Auth options bullet and the closing 'Avoid query-string credentials when header authentication is available' line say the same thing.
Merge the two consecutive Tool Discovery sentences about the live hosted catalog and the published manifest into one statement to trim redundancy.
Add a one-line install/config example for the stdio fallback (e.g., the npx or pip command and the corresponding stdio mcpServers entry) so the Stdio Compatibility section is as executable as the hosted path.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence (no MCP or OAuth tutorials), but query-string credential guidance is stated twice ("supported but not recommended; prefer a header so the key doesn't end up in logs or history" and later "Avoid query-string credentials when header authentication is available"), and the Tool Discovery opener makes the live-catalog point in two consecutive sentences. Minor trims are available, so this is anchor 4 rather than 5. | 4 / 5 |
Actionability | Copy-paste-ready JSON configs for OAuth and API-key auth, the exact CLI command (`claude mcp add --transport http agentmail https://mcp.agentmail.to/mcp`), per-client config file paths, and exact UI navigation sequences for Claude.ai and ChatGPT. The stdio fallback lacks an install command, but anchor 5 only requires the common cases to be fully executable, which they are. | 5 / 5 |
Workflow Clarity | "Verify" is a numbered sequence with an explicit read-only smoke test ("Call `list_inboxes` as a read-only smoke test. Call `list_apps` to confirm the AgentID tools are exposed") plus annotation approval checks, and "Troubleshoot" maps errors to concrete fixes (404 → missing `/mcp`; 401 → key or OAuth recovery), providing validation checkpoints and error-recovery feedback loops per anchor 5. No destructive or batch operations, so the validation cap does not apply. | 5 / 5 |
Progressive Disclosure | The body is well-sectioned and navigable with no nested references, matching anchor 4's "good structure; most content is appropriately placed." It falls short of anchor 5 because the Tool Discovery section inlines catalog-group detail (tool names and workflow descriptions) that could live in a reference file, the sibling skills are named rather than linked, and at ~105 lines the skill exceeds the under-50-line self-contained case the guidelines exempt. | 4 / 5 |
Total | 18 / 20 Passed |