The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to manipulate the subagent's behavior.\n
Ingestion points: Dependency release notes, commit messages, and changelogs are fetched from external sources including NPM, GitHub, and GitLab (SKILL.md).\n
Boundary markers: The skill does not employ delimiters or explicit "ignore embedded instructions" warnings when presenting external content to the subagent.\n
Capability inventory: Subagents are granted shell and network access to run tools like git, npm, gh, glab, and curl (SKILL.md), though they are instructed to remain in a read-only mode.\n
Sanitization: There is no evidence of sanitization or filtering of the fetched external content before it is processed by the model.\n- [EXTERNAL_DOWNLOADS]: The skill fetches documentation and metadata from well-known technology services and registries.\n
Downloads release notes and metadata from github.com, gitlab.com, npmjs.com, and cdn.jsdelivr.net to analyze dependency updates.\n- [COMMAND_EXECUTION]: The skill utilizes several CLI tools to perform its repository and dependency analysis.\n
Uses git, npm, gh, and glab to inspect the local environment, compare package versions, and retrieve security advisories.

dependency-changelog-review