Agent Economy ReportAgent Economy ReportSAFE version 1.0.0 · audited 2026-09-17
The package is a small Markdown skill that documents how to run the Gemini CLI in one-shot mode for questions, summaries, generation, JSON output, and extension management. It declares a Homebrew install for gemini-cli and gives shell command examples. It warns that prompts are processed by Google's Gemini service, that Homebrew installation changes the local machine, and that the --yolo flag should be avoided. No hidden scripts, direct file writes, persistence, wallet access, or credential exfiltration were found. The observed behavior matches the declared purpose.
| Declared purpose | Gemini CLI for one-shot Q&A, summaries, and generation. |
| Observed behavior | The skill is documentation plus metadata only. It instructs an agent or user to run the gemini CLI through shell commands with positional prompts, optional model and output-format flags, extension commands, and a login flow if needed. It advertises Homebrew installation of gemini-cli. The only external domains are ai.google.dev and clawhub.ai, both declared. No executable payload or persistence mechanism is present. |
| Verdict, rules only | SAFE |
| Verdict, AI | SAFE |
| Final verdict | SAFE (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| network | needed | The gemini CLI sends prompts to Google's Gemini service; the skill card warns that prompt content may be processed by Google. |
| shell | needed | The documented examples require running the gemini binary in a shell. |
| install_packages | needed | SKILL.md frontmatter declares a Homebrew formula gemini-cli to provide the gemini binary. |
| credentials | needed | SKILL.md says to complete the Gemini login flow if auth is required; the package itself does not read stored credentials. |
| severity | finding |
|---|---|
| info | Declared purpose The skill describes itself as a Gemini CLI wrapper for one-shot Q&A, summaries, and generation. SKILL.md:3 description: Gemini CLI for one-shot Q&A, summaries, and generation. |
| info | Declared external link Homepage link points to ai.google.dev, a declared Google domain. SKILL.md:4 homepage: https://ai.google.dev/ |
| low | Homebrew install metadata The frontmatter declares that the gemini binary is required and can be installed via the Homebrew formula gemini-cli; this modifies the local environment if applied. SKILL.md:5 metadata: {"clawdbot":{"emoji":"♊️","requires":{"bins":["gemini"]},"install":[{"id":"brew","kind":"brew","formula":"gemini-cli","bins":["gemini"],"label":"Insta |
| info | Usage guidance Tells the agent to avoid interactive mode and use positional prompts. SKILL.md:10 Use Gemini in one-shot mode with a positional prompt (avoid interactive mode). |
| low | Shell command example Demonstrates invoking the Gemini CLI through the shell. SKILL.md:13 - `gemini "Answer this question..."` |
| low | Authentication/login flow The skill may require completing Gemini's interactive login flow if authentication is missing; the package does not directly read credentials. SKILL.md:22 - If auth is required, run `gemini` once interactively and follow the login flow. |
| low | Safety guidance against --yolo Explicitly tells the user or agent not to use the --yolo flag, which could reduce execution safety. SKILL.md:23 - Avoid `--yolo` for safety. |
| info | Publisher link Declared publisher link to clawhub.ai; no hidden destination. skill-card.md:9 [steipete](https://clawhub.ai/user/steipete) |
| low | Google data processing The skill card warns that prompts are processed by Google's Gemini service; users should avoid secrets or private data unless appropriate. skill-card.md:24 Risk: Prompt content may be processed by Google's Gemini service. |
| low | Homebrew environment change The skill card highlights that installing gemini-cli via Homebrew modifies the local environment. skill-card.md:28 Risk: Homebrew installation of the Gemini CLI changes the local environment. |
| low | --yolo risk disclosure The known risk and mitigation state that --yolo can reduce execution safety. skill-card.md:32 Risk: The Gemini CLI --yolo option can reduce execution safety. |
| low | Output type may include shell commands The skill card lists shell commands as a possible output type; the package does not include code to execute model output, but generated shell commands should be reviewed before use. skill-card.md:43 **Output Type(s):** [text, markdown, code, shell commands, configuration, guidance] |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| ai.google.dev | not in the service index | SKILL.md:4 |
| clawhub.ai | not in the service index | skill-card.md:9 |
| Binaries invoked | none |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 1 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 23 | bd1a39d637571c70… |
_meta.json | 6 | 4e745f2770184a5c… |
skill-card.md | 57 | bb773f39fe3eb837… |
For agents
JSON: https://agenteconomy.report/k/gemini.audit.json · badge: https://agenteconomy.report/k/gemini.audited.svg ·
skill rating: /k/gemini · commission an audit of any skill: US$ 29 per version.
The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).