The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes several Python scripts via shell commands (significance-tester.py, performance-monitor.py, campaign-tracker.py). Specifically, the call to performance-monitor.py includes a --data argument containing a JSON object of metrics retrieved from external platforms. If these metrics (such as campaign names or channel labels) contain shell metacharacters or unbalanced quotes, it could result in command injection.
[DYNAMIC_EXECUTION]: The skill executes external logic residing in ${CLAUDE_PLUGIN_ROOT}/scripts/. Because the source code for these scripts is not provided within the skill definition, their full behavior and safety cannot be verified through static analysis of the skill itself.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from various external sources, which creates a potential vector for indirect prompt injection.
Ingestion points: Data is pulled from multiple connected Analytics MCPs (e.g., Google Ads, Meta Marketing, TikTok Ads) and read from local files like profile.json and industry-profiles.md.
Boundary markers: The instructions do not specify any delimiters or safety markers to differentiate between instructions and the external data being analyzed.
Capability inventory: The skill has capabilities including file system access (reading ~/.mcp.json and brand folders) and shell command execution.
Sanitization: There is no evidence of sanitization or validation of the data retrieved from external analytics platforms before it is interpolated into calculations or passed to reporting scripts.
[DATA_EXPOSURE]: The skill accesses sensitive local file paths, including ~/.claude-marketing/brands/ and ~/.mcp.json. While this access is central to the skill's purpose of marketing analysis, these files often contain configuration data and connection details that represent a high-value target if the agent's environment is compromised.

performance-check