Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/aavaz-ai/gmail-mcp/agents-mdgit clone --depth 1 https://github.com/aavaz-ai/gmail-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/aavaz-ai/gmail-mcp/agents-md)<a href="https://agentmods.dev/instructions/aavaz-ai/gmail-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/aavaz-ai/gmail-mcp/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.01137 | $0.01137 |
| Opus 5 | $0.00568 | $0.00568 |
| Sonnet 5 | $0.00227 | $0.00227 |
| Haiku 4.5 | $0.00114 | $0.00114 |
Grade A, and why
gmail-mcp AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Gmail MCP Agent Guidelines
Read this file and README.md before planning or changing the repository.
Workflow
- Default to discovery before implementation: inspect the current repository state, product contract, host approval/retry behavior, and relevant upstream specifications before proposing a code change.
- Preserve existing work. Do not commit, push, create a remote, publish a package, or update external systems without explicit authorization.
- Keep the package small and auditable. Avoid generic Gmail clients, arbitrary REST tools, bundled OAuth frameworks, HTTP transports, token stores, and unrelated Workspace services.
- Update tests and docs with behavior changes. Run the complete credential-free validation before handoff.
Fixed V1 Boundary
The public MCP surface is exactly:
send_emailcreate_labelupdate_labeldelete_label
The only OAuth scopes are:
https://www.googleapis.com/auth/gmail.sendhttps://www.googleapis.com/auth/gmail.labels
Do not add inbox/thread reads, search, drafts, reply/forward, attachments, message labeling or other message mutation, synchronization, triggers, or broader Gmail scopes without a new product and security decision.
Internal provider calls required to resolve label identifiers may not become public read tools or return mailbox/message metadata.
Authentication And Runtime
- The runtime is stdio only and must remain compatible with ephemeral MCP host subprocesses. Deployments should use an exact immutable package version through
npxor a preinstalled reviewed binary. Prefer preinstalled direct execution when bearer material must remain outside package-install processes. - Read one fresh host-supplied bearer from
GMAIL_OAUTH_BEARERat process startup. A missing, blank, or whitespace-only value must fail closed before tools become usable. - Never accept, refresh, or persist refresh tokens. Never write access tokens to disk, caches, snapshots, test fixtures, errors, telemetry, or logs.
- Keep provider error bodies and authorization headers out of user-visible errors. Preserve only safe status/category information needed for diagnosis.
- Reserve stdout for MCP JSON-RPC. Send sanitized diagnostics to stderr.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 73 lines · 1,137 tokens per session scan A 62b011c2f72b
gmail-mcp AGENTS.md is an instructions file published in the GitHub repository aavaz-ai/gmail-mcp (0 stars, last pushed 2d ago), licensed MIT. It adds 1,137 tokens to every session, about $0.0057 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
spellbook AGENTS.md
AGENTS.md instructions for majiayu000/spellbook, covering spellbook agent contract, routing, scope rules, threads long-run guardrails and validation.
codemark GEMINI.md
Gemini CLI instructions for DanielCardonaRojas/codemark, covering development workflow and debugging with the tui logging system.
vrchat-mcp AGENTS.md
AGENTS.md instructions for BASIC-BIT/vrchat-mcp, covering agents, tool ergonomics goals (distilled), curated output philosophy, specific notes (current direction) and caching considerations (planned).
minimalist copilot-instructions.md
Instructions for DivyeshJayswal/minimalist, covering minimalist (coding tasks), persistence, fast path, bias and output.
totem AGENTS.md
AGENTS.md instructions for mmnto-ai/totem, covering totem: agent instructions, what totem is, session start protocol (mandatory), essentials and totem workflow.
front-mcp CLAUDE.md
Claude Code instructions for wearehoust/front-mcp, a project described as: Secure MCP server for Front.com: search conversations, manage contacts, send messages, automate inbox workflows — 26 tools, 172 actions.