The Agent Skills Directory

[SAFE]: The skill performs deterministic analysis of customer data provided in JSON format using local Python scripts. All scoring logic is algorithmic and based on predefined weights and thresholds defined in the reference guides.- [SAFE]: All Python scripts (health_score_calculator.py, churn_risk_analyzer.py, expansion_opportunity_scorer.py) rely exclusively on the Python standard library and do not include external dependencies, network communication, or remote code execution.- [SAFE]: No prompt injection, obfuscation, or persistence mechanisms were identified in the metadata, instructional content, or script logic.- [INDIRECT_PROMPT_INJECTION]: The skill processes customer data from JSON files, creating an ingestion surface. 1. Ingestion points: Scripts in the scripts/ directory ingest data via json.load. 2. Boundary markers: Data is processed as discrete values for mathematical scoring rather than being interpolated into prompts. 3. Capability inventory: The analysis environment has no network, shell execution, or file writing capabilities. 4. Sanitization: Inputs are handled as numeric or categorical values for algorithmic processing, which inherently prevents instruction execution.

customer-success-manager