The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to clone a repository from an untrusted GitHub account.
Evidence: git clone https://github.com/shayantimes/marketing-analytics-lab.git in the Installation section of SKILL.md.
[REMOTE_CODE_EXECUTION]: The skill requires the agent to install dependencies from a downloaded requirements file, which can result in the execution of arbitrary code during the installation process.
Evidence: pip install -r requirements.txt following the repository clone.
[COMMAND_EXECUTION]: The skill provides shell commands for repository cloning, dependency installation, and running tests, which require the agent to have shell access.
Evidence: git clone, pip install, and pytest commands provided in the documentation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests campaign data from external CSV files, which is a common vector for indirect prompt injection if the data contains malicious instructions that the agent might later interpret.
Ingestion points: UnifiedLoader.load(source, filepath) calls pd.read_csv(filepath) to ingest marketing data.
Boundary markers: None identified; data is processed directly after loading.
Capability inventory: The skill performs file reading and provides instructions for shell command execution (git/pip/pytest).
Sanitization: The skill uses Pydantic for schema and type validation (e.g., CampaignRecord), which enforces data structure but does not sanitize string fields against natural language instructions targeted at the LLM.

marketing-analytics-lab-engine