The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill instructions and templates recommend the global installation of the autoship package via pnpm and npx. This package is not from a verified or trusted organization listed in the security policy, posing a potential supply chain risk if the package name is typosquatted or compromised.
[REMOTE_CODE_EXECUTION] (MEDIUM): Documentation in references/ci-integration.md suggests installing the GitHub CLI using a curl | sudo pipe (curl -fsSL ... | sudo dd ...). While targeting a known tool, this pattern circumvents standard package management security and is categorized as high-risk execution. The severity is mitigated by the skill's primary purpose of CI/CD automation.
[COMMAND_EXECUTION] (MEDIUM): The skill makes extensive use of powerful shell commands via the autoship CLI and GitHub CLI to perform sensitive repository operations, including branch creation, merging, and publishing to npm.
[PROMPT_INJECTION] (LOW): The skill exhibits a surface for Indirect Prompt Injection.
Ingestion points: The workflow analyzes git commits and repository diffs to generate release descriptions and suggest version bumps (SKILL.md).
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands within the analyzed commit messages.
Capability inventory: The skill (via autoship) has the capability to create pull requests, merge code, and trigger npm publishing actions.
Sanitization: There is no evidence of sanitization for the commit data before it is processed by the AI.
Risk: An attacker with commit access could craft a malicious commit message designed to influence the AI's versioning suggestions or inject malicious content into the final release notes.

autoship