seo-new-keywords-research
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone the
new-term-radarrepository from GitHub (https://github.com/zhuyansen/new-term-radar) if it is not already present on the local system. - [REMOTE_CODE_EXECUTION]: The skill executes a Python module (
python -m radar) from the externally cloned repository. This pattern involves downloading and running code from a remote source at runtime. - [COMMAND_EXECUTION]: The agent is instructed to run shell commands to manage the repository, configure the environment, and execute the keyword research tool.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (such as listening reports, social media posts, or release notes) to extract keyword names which are then used as arguments in a shell command.
- Ingestion points: Raw text, reports, or posts provided by the user (as described in SKILL.md, Step 1).
- Boundary markers: None identified. The instructions do not specify the use of delimiters to isolate untrusted data when passed to the command line.
- Capability inventory: Shell command execution via
python -m radar(documented in SKILL.md, Step 2). - Sanitization: The instructions suggest converting names to a "searchable form" (lowercase), but there is no explicit sanitization logic to handle or escape shell-sensitive characters that might be present in the processed text.
Audit Metadata