The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands with interpolated user-provided strings. Specifically, parameters like {content_or_path} and {type} are passed directly to python scripts/eval-runner.py. This presents a risk of command injection if the input strings contain shell metacharacters and are not properly escaped by the underlying execution environment.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection as it processes untrusted external data (campaign assets, content libraries) and passes it to an AI-driven evaluation pipeline.
Ingestion points: User-provided file paths, directory contents, and inline content blocks ingested during Step 2.
Boundary markers: There are no explicit delimiters or system instructions defined to prevent the AI from obeying commands hidden within the content being evaluated.
Capability inventory: The skill has the ability to execute subprocesses via scripts/eval-runner.py and scripts/quality-tracker.py and can read sensitive brand profiles from the local file system (~/.claude-marketing/).
Sanitization: The instructions do not specify any validation, escaping, or sanitization of the content before it is processed by the evaluation scripts.

eval-suite