The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface via local file ingestion. The skill performs validation by reading the contents of plugin.json and several markdown files (SKILL.md, agent files, command files) from a user-specified path. Maliciously crafted instructions inside these files could influence the agent during the validation process.
Ingestion points: Reads the content of multiple files within the [plugin-path] directory, including configuration and documentation files.
Boundary markers: The instructions do not define specific delimiters or "ignore previous instructions" guards when reading the content of the files to be validated.
Capability inventory: The skill uses Bash, Read, Glob, and Grep to interact with the filesystem and inspect file contents.
Sanitization: There is no evidence of sanitization or escaping of the content read from files before it is processed by the agent.

validate-plugin