puff
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and installs the
puffCLI utility from the author's personal repositories, including thetkersey/tap/puffHomebrew tap and thetkersey/skills-zigGitHub repository. - [COMMAND_EXECUTION]: The skill executes local shell commands to compile and install the tool, specifically using
zig buildto create the binary from source andinstallto place it in the local path, as well asbrew installfor macOS systems. - [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection. Ingestion points: The
launchandsubmitcommands inSKILL.mdaccept a--promptargument for task instructions, and thejoin-operatorcommand processes data from an external patch inbox. Boundary markers: There are no boundary markers or instructions to disregard embedded commands when these prompts are processed. Capability inventory: The tool can execute tasks in the cloud, monitor background processes, and modify local files via thecodex cloud applycommand. Sanitization: The skill does not perform any validation or sanitization of the input text before it is used in cloud task execution.
Audit Metadata