The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md describe a process for building and executing Bash commands using Python (e.g., python "${CLAUDE_PLUGIN_ROOT}/scripts/eval-runner.py" --action run-quick --file <input>). Because the agent is instructed to interpolate user-provided arguments—such as <file-path-or-content>, --brand <slug>, and --evidence <path>—directly into these shell commands, there is a significant risk of command injection if the inputs are not properly sanitized.
[DATA_EXPOSURE]: The skill is designed to read and list files from the user's home directory at ~/.claude-marketing/brands/. Accessing the ~/ path is a sensitive operation that exposes user-specific data to the agent context. While intended for brand profile management, this capability could be abused to access other sensitive files if the path resolution logic is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content (marketing drafts) through various scripts to detect hallucinations and brand compliance. This creates an attack surface for indirect prompt injection.
Ingestion points: The skill ingests raw content or files via the <file-path-or-content> argument described in SKILL.md.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the content are provided.
Capability inventory: The skill uses the Bash tool to execute Python scripts, performs file system reads, and writes content to temporary files.
Sanitization: There is no evidence of sanitization or filtering of the input content before it is processed by the evaluation suite.

check