The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill guides users to install the 'ecc-agentshield' package from the npm registry. This package originates from a third-party developer ('affaan-m') and is not associated with any recognized trusted organization or well-known service.
[REMOTE_CODE_EXECUTION]: The scan commands rely on 'npx ecc-agentshield', which dynamically downloads and executes remote code at runtime. Executing unverifiable code from non-trusted sources poses a significant risk of arbitrary code execution.
[PROMPT_INJECTION]: The skill acts as a security scanner for local project files such as 'GEMINI.md' and agent instructions. This introduces a vulnerability to indirect prompt injection, where malicious instructions embedded in these files could influence the behavior of the agent during the scan.
Ingestion points: The tool processes content from '.gemini/settings.json', 'GEMINI.md', 'mcp.json', and files in 'hooks/' and 'agents/' directories.
Boundary markers: There are no specified delimiters or instructions to ignore embedded commands in the scanned files.
Capability inventory: The skill environment executes shell commands via 'npx'.
Sanitization: No evidence of sanitization or validation of the scanned content is provided in the documentation.

security-scan