setup-web-tests
Fail
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches a standalone Python binary distribution from the official
astral-sh/python-build-standalonerepository on GitHub. This is a well-known and trusted source for pre-built Python environments. - [COMMAND_EXECUTION]: The skill requires elevated privileges via
sudoto perform the following system modifications: - Appends service hostnames (e.g., kafka, clickhouse) to the system-wide
/etc/hostsfile usingsudo tee. - Installs system-level development libraries (
libxml2-dev,libxmlsec1-dev,libxmlsec1-openssl) viasudo apt-get install. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from the local environment:
- Ingestion points: Reads the
pyproject.tomlfile to extract version constraints. - Boundary markers: No specific delimiters or safety instructions are provided to the agent regarding the content of
pyproject.toml. - Capability inventory: The agent is instructed to perform network downloads (
curl), file extraction (tar), and execute commands with administrative privileges (sudo). - Sanitization: While it uses a regex (
[\d.]+) to extract versions, an attacker who can modifypyproject.tomlcould potentially influence the resulting shell commands or version detection logic.
Recommendations
- AI detected serious security threats
Audit Metadata