better-interface
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is instructed to treat project files like
AGENTS.mdandCLAUDE.mdas 'sources of truth' for conventions. This creates a surface for indirect prompt injection where an attacker could place malicious instructions in these files to influence the agent's behavior or conclusions during a review.\n - Ingestion points: Documentation files including
CONTRIBUTING.md,CODING_STANDARDS.md,AGENTS.md, andCLAUDE.mdare explicitly listed as files to be read and followed inSKILL.md.\n - Boundary markers: The instructions do not define delimiters or protective markers when interpolating the content of these files into the agent's context.\n
- Capability inventory: The skill's capabilities include generating markdown reports and executing local shell commands found in the repository.\n
- Sanitization: No sanitization or validation of documentation content is implemented to prevent instruction override.\n- [COMMAND_EXECUTION]: The skill is directed to 'Identify... any preview or test command' and 'Run the safe, relevant checks the project offers'. This involves the execution of scripts and commands defined within the repository's configuration (such as scripts in
package.json), which could result in the execution of malicious code if the project content is compromised.
Audit Metadata