The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests campaign-level data and metadata from external advertising platforms.
Ingestion points: External spend data and cost metrics retrieved from Google Ads, Meta, LinkedIn, and TikTok.
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from following instructions potentially embedded in campaign names or descriptions.
Capability inventory: The skill executes Python scripts (scripts/ad-budget-pacer.py, scripts/budget-optimizer.py) and performs file write operations to save snapshots, which could be misused if malicious input were processed.
Sanitization: No evidence of data sanitization or validation of the external strings before they are used in analysis scripts.
[COMMAND_EXECUTION]: The skill executes local Python scripts to perform its primary functions.
Evidence: Step-by-step logic involves calling scripts/ad-budget-pacer.py, scripts/budget-optimizer.py, and scripts/performance-monitor.py. These scripts are part of the skill's own package and are used for data analysis and state persistence.

budget-tracker