pypi-security-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and installing various security tools from external sources.
- Fetches the official
uvinstallation script fromastral.sh(Astral). - Downloads the
phylumsecurity CLI fromphylum.io(Phylum). - Retrieves the
dopplerCLI installer fromdoppler.com(Doppler). - References various security tools from
socket.devandosv.dev. - [REMOTE_CODE_EXECUTION]: The documented installation procedures for several tools utilize remote script execution patterns common for CLI utilities.
curl -LsSf https://astral.sh/uv/install.sh | shfor macOS/Linuxuvinstallation.powershell -c "irm https://astral.sh/uv/install.ps1 | iex"for Windowsuvinstallation.curl -sSL https://sh.phylum.io/ | shfor Phylum installation.curl -Ls https://cli.doppler.com/install.sh | shfor Doppler installation.- [COMMAND_EXECUTION]: The skill contains multiple examples of shell commands intended for environment setup, package management, and security auditing.
- Utilizes
pipanduvfor package installation and lockfile management. - Demonstrates usage of
1Password(op) andDopplerfor secure secret injection into the shell environment viaevalandruncommands. - Provides GitHub Actions YAML configurations for automated security scanning and trusted publishing.
- [INDIRECT_PROMPT_INJECTION]: The skill describes workflows for processing untrusted external data, such as PyPI package content and vulnerability database responses.
- Ingestion points: Tools recommended (e.g.,
pip-audit,uv-secure,socket,phylum) ingest data from PyPI, GitHub, and local lockfiles (SKILL.md). - Boundary markers: The skill is a set of instructions; no explicit programmatic boundary markers are defined for the tools' execution logic.
- Capability inventory: The skill provides instructions involving network access (registry queries), file system operations (lockfile reading/writing), and tool execution (
uv,pip,phylum). - Sanitization: The guidance focuses on utilizing mature, established tools designed to safely process package metadata and SBOM files.
Audit Metadata