TinyFish Web Agent

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

ConcernMedium Confidence
ASI02: Tool Misuse and Exploitation
What this means

The agent could be used to automate or scrape sites that are trying to restrict bots, creating legal, account, or abuse risk for the user.

Why it was flagged

The skill explicitly authorizes scraping and automation against bot-protected sites, which is broader than normal page extraction and lacks artifact guidance limiting use to authorized targets.

Skill content
Use when you need to extract/scrape data from websites, handle bot-protected sites, or automate web tasks.
Recommendation

Use only on sites you are authorized to automate, and require explicit user approval before any non-read-only browser action.

What this means

A vague or mistaken goal could cause unwanted clicks, form submissions, or changes on third-party websites.

Why it was flagged

The documented browser agent can interact with arbitrary websites and fill forms, but the skill does not include safety instructions to stop before submissions, purchases, posts, or account-changing actions.

Skill content
The agent opens a real browser, navigates, clicks, fills forms, and extracts data.
Recommendation

Keep tasks read-only by default and require confirmation before submitting forms, logging in, purchasing, posting, or changing account data.

What this means

Installing a global CLI gives third-party package code local execution on the user's machine.

Why it was flagged

The skill is instruction-only but tells the user/agent to install a global npm package that is not represented in an install spec or pinned version.

Skill content
If not installed: `npm install -g @tiny-fish/cli`
Recommendation

Verify the npm package source and publisher before installing, and prefer a pinned version or isolated environment.

What this means

The TinyFish CLI will operate under the user's TinyFish account and may incur provider usage or expose tasks to that account.

Why it was flagged

The skill requires TinyFish account authentication even though registry metadata lists no primary credential or required environment variable; this is expected for the provider but under-declared.

Skill content
If not authenticated: `tinyfish auth login` or set `TINYFISH_API_KEY` env var.
Recommendation

Use a least-privileged API key where available, keep it out of prompts and logs, and revoke it if no longer needed.

What this means

Sensitive URLs, prompts, or form data included in automation goals could be visible to the external provider.

Why it was flagged

The skill routes URLs and extraction tasks through TinyFish's server-side service, so submitted URLs, goals, and extracted results may be processed outside the local agent.

Skill content
Accepts multiple URLs in a single call — they are fetched in parallel server-side
Recommendation

Avoid sending private or regulated data unless TinyFish's data handling terms meet your requirements.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

Automation jobs may keep running or consuming credits unless monitored and canceled when needed.

Why it was flagged

The skill supports asynchronous and batch agent runs that can continue after submission, though it also documents list/get/cancel commands.

Skill content
`--async` | Submit and return immediately ... `tinyfish agent batch run --input runs.csv`
Recommendation

Use async and batch modes only for well-scoped jobs, then check run status and cancel unintended runs promptly.