The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (!command) to execute shell commands at load time for project introspection.
Evidence: SKILL.md executes a Node.js script to parse package.json for dependencies and a find command to locate configuration files (lines 22-25).
These commands are used solely for identifying the project structure and do not involve network operations or sensitive file access.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from the local filesystem into the agent context.
Ingestion points: SKILL.md reads and outputs content from package.json using the node -e dynamic context command.
Boundary markers: No clear delimiters or safety instructions are present to prevent the agent from following directions potentially embedded in the ingested project metadata.
Capability inventory: The skill has access to Bash (limited to node -e and find) as defined in the allowed-tools frontmatter.
Sanitization: No validation or sanitization is performed on the data extracted from package.json before it is placed in the context.

framework-to-capacitor