install-skill-pack

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill downloads and executes the 'skills' CLI tool from the official npm registry using the npx command during the installation and verification phases.
  • Evidence: Instructions in references/skills-cli-usage.md for resolving the version via npm view and executing via npx --yes "skills@$version".
  • [COMMAND_EXECUTION]: The skill utilizes the host's shell to execute various Git operations, including cloning repositories, fetching specific commits, and checking out detached heads.
  • Evidence: Shell commands in references/revision-selection.md such as git clone, git ls-remote, git fetch, and git checkout.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, user-provided Git repositories, which serves as a primary attack surface for instructions hidden in external content.
  • Ingestion points: The skill clones remote repositories and reads all files within them, specifically targeting SKILL.md and referenced scripts or assets for review and installation (SKILL.md Step 5, references/revision-selection.md).
  • Boundary markers: The skill requires resolving a specific reviewed_sha and performing a manual review step before proceeding with installation (SKILL.md Step 6, references/revision-selection.md).
  • Capability inventory: The skill can execute shell commands via git, npm, and npx, and it modifies global configuration directories and creates symbolic links in the user's home directory (references/skills-cli-usage.md).
  • Sanitization: The skill relies on instruction-based review criteria to detect path escapes, obfuscation, and unauthorized network access, as defined in references/source-review.md.
  • [DYNAMIC_EXECUTION]: The skill dynamically resolves the version of a remote package at runtime and executes it, which can be manipulated if the registry or the version resolution process is compromised.
  • Evidence: The workflow in references/skills-cli-usage.md that checks skills@latest and then uses the resulting version string to invoke npx.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 12:17 PM
Security Audit — agent-trust-hub — install-skill-pack