google-tools-mcp: Skill for Claude Code

.claude/skills/live-agent-loop/SKILL.md

live-agent-loop is a skill for Claude Code from karthikcsq/google-tools-mcp. It costs 0 tokens per session (2,522 once invoked), scanned A, original, MIT.

A live test process for checking whether an AI agent can use Google API tools to complete a real task. It gives the agent a goal, observes where it struggles, and treats confusing tool behavior as a defect.

In plain words
What is it for?
It helps test Google API tool workflows against a real account, collect friction reports, and improve the tools until an independent agent can complete the mission. Writes are restricted to a specified sandbox folder and Gmail sending is blocked.
Why use it?
Unit tests check whether code behaves as expected in prepared cases, but they do not show whether an agent can discover and use the tools without guessing or workarounds.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions subagents.

This is karthikcsq/google-tools-mcp's own configuration. It tells Claude Code how to work on google-tools-mcp itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything google-tools-mcp configures →

Reuse

Borrowing it

Nothing to install: this file belongs to karthikcsq/google-tools-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/karthikcsq/google-tools-mcp/main/.claude/skills/live-agent-loop/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/karthikcsq/google-tools-mcp

Made for: Claude Code.

Wrote 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.

agentmods badge for live-agent-loop

README.md
[![agentmods](https://agentmods.dev/badge/skills/karthikcsq/google-tools-mcp/live-agent-loop/github.svg)](https://agentmods.dev/skills/karthikcsq/google-tools-mcp/live-agent-loop)
Your own site
<a href="https://agentmods.dev/skills/karthikcsq/google-tools-mcp/live-agent-loop"><img src="https://agentmods.dev/badge/skills/karthikcsq/google-tools-mcp/live-agent-loop/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for live-agent-loop

Your own site · 80×15
<a href="https://agentmods.dev/skills/karthikcsq/google-tools-mcp/live-agent-loop"><img src="https://agentmods.dev/badge/skills/karthikcsq/google-tools-mcp/live-agent-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,522 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.02522
Opus 5 $0.00000 $0.01261
Sonnet 5 $0.00000 $0.00504
Haiku 4.5 $0.00000 $0.00252

Measured 7d ago against content hash 4ca95a093011, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

live-agent-loop 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 7d 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.

.claude/skills/live-agent-loop/SKILL.md · 223 lines

How it starts

The opening of the file, as written. The whole thing — 223 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Live agent loop

Unit tests prove a tool does what its author expected. They cannot prove a tool is usable by an agent that was handed a goal and no instructions. Those are different questions, and only the second one predicts what happens after release.

This skill runs the second test: give a subagent a real objective, make it reach the objective using only the MCP tools against the real Google API, and treat everything it struggled with as a defect.

The core rule

Friction is a bug. If the agent needed three attempts, guessed a parameter name, had to read the source, or worked around an error message that did not say what was wrong, that counts as a failure even though the task eventually succeeded. A future agent will hit the same wall with less patience.

Do not fix the mission to avoid the friction. Fix the tool.

Safety, before anything else

Every run touches a real Google account. The envelope is non-negotiable:

  • Every write is confined to the sandbox Drive folder named by GOOGLE_MCP_TEST_FOLDER_ID in .env.live-smoke. There is no default and no fallback; the runner refuses to start without it.
  • Gmail send paths are blocked at the harness level. sendMessage, sendDraft and friends refuse in BLOCKED_TOOLS.
  • Containment is re-verified at cleanup time, so a run trashes exactly what it created inside the sandbox and nothing else.

Never disable, weaken, or route around any check in scripts/live-smoke/guard.mjs. If the guard blocks something you believe is legitimate, that is a finding to report, not an obstacle to remove.

The three harnesses, and when each applies

Runner Scope Use it for
npm run live-smoke checked-in scenarios with assertions, live/ regression: things we already decided to assert
npm run live-call one tool, one process a quick probe of a single tool you just changed
npm run live-mission one multi-step task, one process this loop: goal-driven work with real session state
npm run live-coverage no network; static which of the 160 tools live code actually drives, and which it does not

Read the full file on GitHub · 223 lines

Changes

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.

  1. 7d ago First seen · 223 lines · 0 tokens per session scan A 4ca95a093011

Subscribe to this mod's changes

live-agent-loop is a skill published in the GitHub repository karthikcsq/google-tools-mcp (9 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,522 tokens. 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 2026-09-05.

Related

Other skills, from other repositories

llm-as-judge-evaluation

Evaluate LLM outputs using frontier models as judges. Use for pairwise model comparison, quality scoring with custom rubrics, and automated evaluation pipelines. Covers position bias mitigation, statistical significance, and generating preference data for DPO/RLHF.

synthetic-sciences/openscience · 56 tokens

langsmith-observability

LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.

synthetic-sciences/openscience · 45 tokens

impl-validator

Validate whether an implementation matches its stated goal. Use this skill when a skill or agent wants a second opinion on its own output, when the user says "check this implementation", "validate what you did", "is this correct?", "review the output", or "did you do this right?". Also spawned automatically as a…

Ar9av/obsidian-wiki · 105 tokens

setup

Set up or update TDD Guard for the current project. Detects the test framework, installs or updates the matching reporter, and configures or migrates its configuration to match the current specification.

nizos/tdd-guard · 39 tokens

managing-google-workspace

Manages Google Workspace operations across 12 services (Gmail, Drive, Calendar, Docs, Sheets, Slides, Forms, Tasks, Contacts, Chat, Apps Script, Custom Search). Supports MCP tools or CLI via uvx workspace-mcp --cli. Provides tool routing, workflows, and parameter guidance for 114 tools. Triggers for "check my email"…

taylorwilsdon/google_workspace_mcp · 123 tokens

semgrep-rule-variant-creator

Creates language variants of existing Semgrep rules. Use when porting a Semgrep rule to specified target languages. Takes an existing rule and target languages as input, produces independent rule+test directories for each language.

waybarrios/opencode-power-pack · 50 tokens