Gini Agent is a personal AI-agent runtime that manages conversations, runs, tasks, approvals, memory, skills, jobs, tools, traces, audit events, and runtime health across its clients. It is for an agent that can remember, learn from task outcomes, and operate through desktop or mobile interfaces with user controls. Catalogue add-ons extend the runtime with skills and instructions.
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 skills/open-curiosity/gini-agent/google-gmailnpx skills add Open-Curiosity/gini-agent --skill google-gmailgit clone --depth 1 https://github.com/Open-Curiosity/gini-agentWrote 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/skills/open-curiosity/gini-agent/google-gmail)<a href="https://agentmods.dev/skills/open-curiosity/gini-agent/google-gmail"><img src="https://agentmods.dev/badge/skills/open-curiosity/gini-agent/google-gmail.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 | $0.00024 | $0.04177 |
| Opus 5 | $0.00012 | $0.02089 |
| Sonnet 5 | $0.00005 | $0.00835 |
| Haiku 4.5 | $0.00002 | $0.00418 |
Grade B, and why
google-gmail scanned grade B with 1 finding 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 5d 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.
Subtle steeringmediumPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
After you save a draft (`--draft`), show it to the user inline so they can read it right in the chat — never tell them to open Gmail and search for it. Lead with one short sentence, then render the draft as a fenced `ema How it starts
The opening of the file, as written. The whole thing — 232 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Google Gmail
Use gws gmail to read, search, send, reply, forward, draft, label, and triage Gmail directly from the terminal. The CLI wraps the Gmail v1 API and produces structured JSON, so it composes cleanly with jq and other shell tooling.
Parsing gws output
gws writes its JSON to stdout and a Using keyring backend: keyring preamble (plus any warnings) to stderr. terminal_exec already shows the two streams as separate blocks, so on its own the preamble is harmless — but the moment you pipe gws into a JSON parser you must strip stderr first, or the preamble lands on the JSON and the parse throws:
gws ... 2>/dev/null | jq ... # correct: stderr dropped before the pipe
gws ... 2>/dev/null | python3 -c 'import sys,json; json.load(sys.stdin)'
Never use 2>&1 when piping into a parser — it folds the preamble onto the JSON and breaks it. Do not pass --format text either; it is invalid (valid formats are json, table, yaml, csv), and the raw users.* API already defaults to JSON. When you just need the data and don't need to parse it yourself, prefer the curated +helpers, which print clean output. Note that 2>/dev/null also discards gws's own error messages, so if a command returns empty or unexpected output, re-run it without the redirect (or check the exit code) to see the actual error.
Prerequisites
- If this deployment is managed/hosted, the Google credential is already provisioned at sign-in —
gwsis installed and authenticated, so skip the setup flow below and rungwsdirectly. Scopes are fixed at sign-in on a managed deployment: when a call fails withscope requiredor HTTP 401, tell the user which action needs a scope their account wasn't granted, instead of trying to set anything up. (The scope list below still describes which verb needs which scope.) gwsinstalled and authenticated. Ifgwsis not on PATH ORgws auth statusreports no authenticated user, do NOT silently call setup. Instead, in a single short reply to the user:- State plainly what's missing — e.g. "Google Workspace access isn't set up on this machine yet" or "your Google sign-in has expired."
- Ask one sentence: "Want me to walk you through setting it up?" Wait for the user's answer.
- If they say yes, call
read_skillwith namegoogle-workspace-setupand run that skill's onboarding flow turn-by-turn. If they say no or ask to defer, acknowledge briefly and stop — do not retry the original request.
- Apply the same flow when any
gws gmail ...call fails mid-task withcommand not found/ ENOENT, HTTP 401, "no credentials", or "scope required". Don't report the failure as a dead end — surface the missing prerequisite and ask if the user wants to set it up before moving on. - The OAuth scopes the user picked at login must cover the verbs the agent will use:
- Read-only triage:
gmail.readonly - Send a new message:
gmail.send - Reply, reply-all, forward:
gmail.modify— upstream helpers fetch the original message to threadIn-Reply-To/Referencesheaders, whichgmail.sendalone cannot do - Drafts and labels:
gmail.modify(orhttps://mail.google.com/for full access including permanent delete) - Watch for new mail (
+watch):gmail.modifyANDhttps://www.googleapis.com/auth/pubsub— Cloud Pub/Sub is a separate Google API and its scope must be granted alongside the Gmail scope
- Read-only triage:
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.
- 5d ago First seen · 232 lines · 24 tokens per session scan B e0b6e00e2833
google-gmail is a skill published in the GitHub repository Open-Curiosity/gini-agent (1,869 stars, last pushed 1mo ago), licensed MIT. It adds 24 tokens to every session and 4,177 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (subtle steering). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
gno
Search local documents, files, notes, and knowledge bases. Index directories, search with BM25/vector/hybrid, get AI answers with citations. Use when user wants to search files, find documents, query notes, look up information in local folders, index a directory, set up document search, build a knowledge base, needs…
vibelution-continuous-optimization-loop
Runs one evidence-driven Vibelution optimization iteration from the ROI backlog: pick one todo item, worktree isolate, implement, machine-verify, merge, update STATE. Use when the user starts /loop for project improvement, continuous optimization, or agent-dev ROI backlog work.
personal-assistant
Use this skill whenever the user acts like Bamboo is their personal assistant. Triggers include stating commitments, deadlines, appointments, or recurring routines ("remind me to...", "I need to... by Friday", "every Monday I..."), asking to be reminded of something, agenda questions ("what's on my plate", "what's due…
review
Review code changes, pull requests, patches, or a scoped code area for actionable correctness, security, compatibility, and test risks with file and line evidence. Use for review or audit requests; do not use for general proofreading, feature implementation, or debugging a reported failure when the user wants a fix.
debug
Diagnose a concrete failure, regression, crash, hang, flaky test, or incorrect runtime behavior by reproducing it, testing hypotheses, and identifying the evidence-backed root cause. Use when symptoms or failing output exist; do not use for feature implementation without a failure, general code review, or a conceptual…
plan
Build an executable implementation, migration, or rollout plan from read-only repository exploration, including constraints, dependencies, ordered changes, verification, risks, and rollback points. Use when the user asks for a plan or decomposition before implementation; do not use when they want a small change…