github-code-search
Fail
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill fetches an installation script from the developer's GitHub repository and executes it by piping the content directly to a shell interpreter (sh on Unix-like systems, iex in PowerShell).
- [EXTERNAL_DOWNLOADS]: Downloads the
grep-app-cliinstaller and its associated binary components from GitHub Releases. - [COMMAND_EXECUTION]: Executes the
grep-app-clitool locally to perform searches based on user-provided queries and filters. - [PROMPT_INJECTION]: The skill processes untrusted code content from external GitHub repositories, presenting a surface for indirect prompt injection.
- Ingestion points: Search results containing arbitrary code snippets and comments from public repositories are ingested into the agent context (SKILL.md).
- Boundary markers: The skill includes explicit instructions in the 'Notes' section telling the agent to treat search results as data only and to never execute or follow instructions found inside the matched code.
- Capability inventory: The skill is designed to perform subprocess execution of the
grep-app-clitool and display results; no autonomous file-writing or further execution capabilities are granted to the results. - Sanitization: No programmatic sanitization or filtering of the external code content is performed before it is processed by the agent.
Recommendations
- HIGH: Downloads and executes remote code from: https://github.com/sudosubin/grep-app-cli/releases/latest/download/grep-app-cli-installer.sh - DO NOT USE without thorough review
Audit Metadata