The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local script named growth-loop-modeler.py with subcommands detect-loops and project. This execution relies on parameters and data derived from untrusted user input.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data that is used by agents and passed to execution tools.
Ingestion points: User-provided Business metrics, Business model, and Known growth drivers are processed to calibrate the model.
Boundary markers: No delimiters or isolation instructions are present to prevent the agent from obeying instructions embedded in the provided data.
Capability inventory: File system read access for brand profiles in ~/.claude-marketing/ and subprocess execution via growth-loop-modeler.py.
Sanitization: No sanitization or validation logic is specified for the external data before it enters the modeling process.

loop-detect