defend-my-project
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands such as
git clone,git grep, andgit logto perform deep analysis of project repositories. It clones user-provided URLs into temporary system directories to inspect source files and commit history. - [DATA_EXPOSURE]: The skill is instructed to specifically search for and identify sensitive credentials including API keys, tokens, and private keys (e.g., sk-..., AKIA...) within the project files and git logs. Although this is intended to help the user identify security leaks, it involves the agent actively identifying and reading potentially sensitive information from untrusted sources.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests and processes untrusted code from external repositories to generate its 'attack list' and defense sheet. 1. Ingestion points: External code repositories cloned via URL or read from the local directory. 2. Boundary markers: None are specified to isolate the untrusted code from the agent's instructional context. 3. Capability inventory: The agent can execute shell commands (git), write files to the
$CRACKED_DIRenvironment, and interact with other internal tools. 4. Sanitization: There is no mention of sanitizing or validating the code content before it is used to influence the agent's logic for generating interview questions. - [EXTERNAL_DOWNLOADS]: The skill performs external network operations by cloning repositories from GitHub. While GitHub is a well-known platform, the content being downloaded is not restricted to trusted sources and is determined by the URL provided during execution.
Audit Metadata