kill-ai-slop
Warn
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The scanner script
scripts/scan.mjsuses the dynamicimport()function to load external JavaScript modules provided via the--rulescommand-line argument. This allows the tool to be extended with custom patterns but also enables the execution of arbitrary JavaScript modules from the local filesystem. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the project being scanned (HTML, CSS, JS, and Markdown files). Because the agent is instructed to triage these findings and apply automated fixes, a malicious actor could embed instructions within these source files to influence the agent's behavior during the 'Report' or 'Fix' phases.
- Ingestion points: Project source files read by
scripts/scan.mjsduring the walk operation. - Boundary markers: None implemented; the agent is instructed to 'open the file and decide slop vs intentional'.
- Capability inventory: Local shell execution (node), file system read, and file system write (diff application).
- Sanitization: The scanner script sanitizes terminal escape sequences in its output but does not sanitize the content processed by the agent.
- [EXTERNAL_DOWNLOADS]: The documentation references an external GitHub repository (
github.com/yetone/kill-ai-slop) as the source for the skill's code and updates. - [COMMAND_EXECUTION]: The workflow requires the agent to execute a bundled Node.js script (
scripts/scan.mjs) to perform its primary function. While intended, this grants the agent execution capabilities within the user's project environment.
Audit Metadata