The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (the ! syntax) to execute shell commands at load time for project discovery. These commands are benign:
A node script reads the local package.json to extract project name, version, and Capacitor dependencies.
A find command lists specific native project directories.
Neither command performs network operations, accesses sensitive credentials, or accepts unsanitized user arguments.
[PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection because it incorporates data from package.json into the agent's context.
Ingestion points: @capacitor/ dependency strings and project metadata from package.json are read into the skill context via dynamic injection in SKILL.md.
Boundary markers: None present.
Capability inventory: The skill allows the agent to run npm install and modify local files.
Sanitization: None.
This finding is assessed as safe because the behavior is standard for a developer tool performing project analysis.
[SAFE]: The skill's metadata and external references point to legitimate Capacitor documentation, and no persistence, exfiltration, or obfuscation patterns were detected.

capacitor-plugin-upgrade-v6-to-v7