Browser Use · Magpie

Browser Use · Magpie

Safety checkRisky, SkillSpector, 19 Oct

Its AI check confirmed something serious. Don’t install it without reading what it found.

Finding (high, rule SDI-4)
skills/x402/SKILL.md:131
The Path C wallet-generation snippet prints the new wallet's private key to stdout (`print("Private key:", acc.key.hex())`, and the TypeScript equivalent). The agent running the skill sees that output, so the key lands in the tool-output transcript and any logs or screenshots. Anyone with access to those can take control of the wallet and its USDC. The skill's own text says to show only the address and never print the key, so the code contradicts the instruction.
Finding (high, rule SQP-2)
skills/x402/SKILL.md:132
Same defect as SDI-4, reported again from the Path C code block. The generated private key is printed to stdout, which exposes a fund-holding key in agent transcripts and logs, against the skill's stated rule.
External Script Fetching (high, rule SC2)
skills/cloud/references/quickstart.md:162
This is the same remote pipe-to-shell instruction as SQP-2, which fetches and runs an unverified script from a vendor domain with no warning. Since the script is tied to browser login data, a compromised or altered script would run with the user's privileges and could access credentials.
External Script Fetching (high, rule SC2)
skills/cloud/references/sessions.md:92
This is the same remote pipe-to-shell pattern as the SQP-2 finding at this line. The domain appears to belong to the vendor, but the pattern still lets remote content run unverified on the user's machine.
External Script Fetching (high, rule SC2)
skills/cloud/references/sessions.md:103
The Authentication Strategies section repeats the unverified `curl | sh` execution, so users are steered into running remote code as the recommended first option. That raises the chance they run it without checking.
External Script Fetching (high, rule SC2)
skills/open-source/references/quickstart.md:147
This is the same risk as the piped remote script above, flagged under the external-script rule. Fetching and executing code from a URL without integrity checks makes the user's environment depend on the server's current content.
Chaining Abuse (high, rule TM2)
skills/open-source/references/quickstart.md:147
Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.
Not confirmed by the AI check.
External Script Fetching (high, rule SC2)
skills/qa/SKILL.md:57
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
Not confirmed by the AI check.

SkillSpector’s own score is 100/100 from 121 findings; only the ones its AI check confirmed count here.