The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute npm view and npm audit commands. These commands incorporate package names extracted directly from local package.json files without explicit validation. If a dependency name in a project file is maliciously crafted (e.g., using shell metacharacters), it could lead to arbitrary command execution on the host system.
[DATA_EXFILTRATION]: To perform its analysis, the skill transmits metadata about the project's dependencies (package names and versions) to external endpoints including the npm registry (registry.npmjs.org), GitHub's API (api.github.com), and general web search engines. This is the primary function of the tool but represents a flow of project configuration data to external entities.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of untrusted external content.
Ingestion points: Data is ingested from local package.json files, responses from the npm registry, GitHub release notes, and arbitrary web search results.
Boundary markers: The instructions do not define boundary markers or delimiters to separate this external untrusted content from the agent's core instructions.
Capability inventory: The agent has access to powerful tools including Bash (shell access) and Write (filesystem modification).
Sanitization: No sanitization, escaping, or validation steps are defined for the content retrieved from remote changelogs or migration guides before it is processed and summarized.

analyze-deps-guide