MCP Safety Board · server report · November 2026
a-bonus/google-docs-mcp
F
50 / 100
Grade F · ★ 686 · TypeScript checked by hand
How the score was worked out
| Check | What was found | Points |
|---|---|---|
| Risky tools labelled | 29 of 29 tools whose name or description suggests delete, write or run declare neither destructiveHint nor readOnlyHint. −2 each (capped at −30). | −30 |
| Tools with any label | 0 of 128 tools (0%) carry any annotation; fewer than half. | −10 |
| README install command | The README tells users to launch the server with npx / bunx / uvx / pnpm dlx without a pinned version (or with @latest). | −10 |
| Committed config | No committed MCP config with an unpinned package, a moving tag, a broad filesystem root, a privileged docker server or plain http. | 0 |
| Score | 100 minus the penalties above (never below 0). | 50 |
- MCP106 (12): A destructive tool has no annotations (recorded only: already counted in the tool-label check). No points.
Changes that would raise this grade
- +30Label the 29 tools that delete, write or run: add destructiveHint (or readOnlyHint) to their annotations. Alone, this gives 80 (B).
- +10Add annotations to 64 more tools so at least half of all 128 tools are labelled (labelling the risky tools above counts too). Alone, this gives 60 (D).
- +10Pin the version in the README install command (pkg@1.2.3 or pkg==1.2.3, not @latest). Alone, this gives 60 (D).
Doing all of these gives 100 (A).
What was checked
- Commit checked
22856e1fd5d2- Repository
- github.com/a-bonus/google-docs-mcp
- Tools found
- 128, of which 0 carry any label (0%)
- Risky tools
- 29, of which 0 labelled
- README install
- not pinned
- Last push
- 2026-10-26
Static reading of this one commit. No code was installed or run. This is a hygiene grade, not a security audit.
Maintainer?
Fixed something, or think a count is wrong? Request a re-check (opens a GitHub issue). Re-checks run on the latest commit.
README badge
[](https://mcp-safety-board.pages.dev/s/a-bonus--google-docs-mcp.html)Want a written review of your agent's tools and permissions?
48-hour written review. Which tools your agent can call, what each one can change, which need a confirmation step, and what to lock down first. Send a message to start.