The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command: node ../../../.fullstack-forge/runtime/cli/src/composition-entry.js jobs compose --root <repository-root> --json. This involves executing a script from a relative path that reaches outside the skill's assumed directory structure, creating a risk if the target script is substituted or the path is manipulated.
[REMOTE_CODE_EXECUTION]: By directing the agent to run composition-entry.js, the skill executes code not contained within its own package. This pattern of executing external logic from the filesystem is a significant security surface.
[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by loading specialist guidance from paths determined at runtime.
Ingestion points: The agent reads .forge/composition.json to identify and load instruction files from the selected paths.
Boundary markers: No boundary markers or warnings to ignore embedded instructions are provided.
Capability inventory: The skill possesses the capability to read local files, execute Node.js scripts, and follow instructions from dynamically identified paths.
Sanitization: No sanitization or validation of the contents of the configuration file or the loaded instruction files is mentioned.

forge-jobs