Gemini CLI by Google

# Fix list: Gemini CLI

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/gemini-cli, the October 2026 research run, assessed 1 October 2026. Grade BB, 72.3 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Gemini CLI: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: Harness reading of the published rubric. No payment protocol (0). Free quotas, plan quotas and API token prices are public without a login (20). 1,000 requests a day with a Google sign-in and no card (20). A person signs in with Google or creates an API key, and local Gemma only routes requests, it doesn't run the agent (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Reliability, 71 out of 100, up to 5.8 more on the total

Why it scored 71: Local-package reading. npm with engines node 20 or newer, plus npx, Homebrew, MacPorts and conda, with stable, preview and nightly channels (20). Public CI, and the 10 most recent Testing CI runs on main all passed, with a chained end-to-end workflow besides (25). 583 open issues and 252 open pull requests, triaged by bot and by hand with priority labels, though several P1 security reports from 13 September were unassigned (18). releases.md says the project follows semver as closely as possible with weekly minors and patch fixes between, but release notes have no breaking-change heading (8). 0.62.0, pre-1.0 (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 3. Security & auth, 67 out of 100, up to 5.8 more on the total

Why it scored 67: Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. Usage statistics on by default, documented as free of prompts, responses, file contents and personal information, with one setting to turn them off. On the free Code Assist for individuals tier Google may use data to improve its models unless the user opts out. Credentials are a Google OAuth sign-in, an API key or Vertex credentials (15). Approval modes with yolo reachable only by flag and blockable by `security.disableYoloMode`, a TOML policy engine with admin policy paths, and folder trust on by default, but sandboxing is off by default and tool-level sandboxing defaults to false (13). Environment-variable redaction, and 0.61.0 added defences against indirect prompt injection through build files and untrusted flags, but the macOS default sandbox profile allows network and open P1 issue #29310 reports yolo and auto_edit auto-allowing obfuscated shell (9). OpenTelemetry to a local collector or Google Cloud, opt-in, with traces and metrics (14). SECURITY.md routes reports to g.co/vulnz with a five-working-day response and google.com has a valid security.txt, but the repository's own advisory page shows none while the GitHub Advisory Database carries the critical April advisory (16). SOC 2 isn't scored on the framework reading.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 4. Agent ergonomics, 78 out of 100, up to 3.6 more on the total

Why it scored 78: Framework reading, adapted to a harness driven by a pipeline. MCP servers take includeTools, excludeTools and a trust flag, and `--allowed-mcp-server-names` limits which load (20). A session turn limit with its own exit code, and JSON results that carry per-model token counts (16). Documented exit codes and an error object in JSON output (18). `--resume` and checkpointing (14). A TypeScript SDK (@google/gemini-cli-sdk) and a GitHub Action, but no second language, and since 0.39.1 a CI run has to set `GEMINI_TRUST_WORKSPACE` to work in its own checkout (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Schema & documentation, 93 out of 100, up to 1.1 more on the total

Why it scored 93: Framework reading. settings.schema.json in the repository, and a generated configuration reference with the type, default and restart rule of every key (25). llms.txt at geminicli.com serving the documentation as Markdown (10). Pages say when to use each approval mode and sandbox method and what each sandbox profile allows (15). Approval modes, sandbox commands and Seatbelt profiles are enums (13). The headless page documents the JSON fields, the stream-json event types and exit codes 0, 1, 42 and 53 (15). A changelog page per stable and preview release, dated, under a written release policy (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Maintenance & community, 88 out of 100, up to 1.1 more on the total

Why it scored 88: 0.62.0 on 2026-09-29 (30). 15 stable releases since 3 July, plus weekly previews and nightlies (20). 583 open issues with priority labels and bot triage, and 252 open pull requests, though some P1 security reports sat unassigned (18). A TypeScript SDK in the repository and the run-gemini-cli GitHub Action, both current, but no SDK in a second language (10). CI, nightly evals and an Allstar policy (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 7. Transparency & trust, 90 out of 100, up to 0.9 more on the total

Made of editorial 80, provenance 100.

Why it scored 90: Apache-2.0 (30). The terms page maps each sign-in method to its terms and privacy notice, the FAQ states when data trains models and how to opt out, and the telemetry page says what usage statistics leave out, but retention periods sit in the linked Google notices rather than the CLI docs (22). A written release policy that promises to call out departures from semver, but no deprecation notices with dates (10). Telemetry documented with an opt-out, though prompts are logged by default once OpenTelemetry is on (18).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-04-24. GHSA-wpqr-6v78-jr5g, critical (CVSS 10). In CI, headless Gemini CLI trusted the workspace folder automatically and loaded its configuration, and `--yolo` ignored fine-grained tool allowlists, so a workflow fed untrusted pull requests or issues could run an attacker's code. Fixed in @google/gemini-cli 0.39.1 and run-gemini-cli 0.1.22 and published, so the deduction is small (https://github.com/advisories/GHSA-wpqr-6v78-jr5g)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- The GHSA-wpqr-6v78-jr5g advisory is in the GitHub Advisory Database, but the gemini-cli repository's own advisory page says there are none, so we couldn't tell which repository published it
- unchecked: whether @google/gemini-cli-sdk is published to npm as a stable package
- unchecked: the incident history of the services behind a Google sign-in
- docs/changelogs/latest.md in the repository still described 0.61.0 when 0.62.0 was tagged on 29 September

## Weaknesses

- Sandboxing is off by default, and the default macOS profile allows network
- Usage statistics on by default, and the free tier may train on data unless the user opts out
- A critical advisory in April 2026 (CVSS 10) for CI runs that trusted untrusted repositories
- Pre-1.0 at 0.62.0, and Gemini models only
- Open P1 report #29310 says yolo and auto_edit auto-allow obfuscated shell commands

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Set `GEMINI_TRUST_WORKSPACE` to true only for trusted inputs in CI. Since 0.39.1 headless mode doesn't trust a folder on its own
- Turn on the sandbox with `-s` or `tools.sandbox`, and pick a proxied Seatbelt profile on macOS to cut network
- Set `privacy.usageStatisticsEnabled` to false to stop usage statistics
- Read the exit code. 42 is bad input and 53 is the turn limit
- Use `--output-format stream-json` to get tool calls and results as JSONL events

## What the review panel asked for

- breaking-change section in notes
- dated deprecation notices
- sandbox on by default
- advisories in the repository

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.