The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through the ingestion of external data. The script scripts/campaign_tracker.py processes campaign data from JSON files that could contain instructions designed to manipulate the agent's behavior. \n- Ingestion points: The scripts/campaign_tracker.py script (line 123) reads campaign data from a user-provided file path using argparse. \n- Boundary markers: The output generated by the script lacks clear delimiters or warnings that would instruct the agent to ignore any potential instructions embedded within the task descriptions or campaign names. \n- Capability inventory: The agent is instructed to orchestrate campaigns and route marketing tasks based on the analysis provided by this skill and its associated script. \n- Sanitization: There is no evidence of string sanitization or validation performed on the JSON content before it is printed to the console for the agent to read.

marketing-ops