The Agent Skills Directory

[SAFE]: The skill performs read-only analysis of a local codebase to identify donor management patterns. It does not attempt to execute remote scripts, download external dependencies, or bypass security constraints.
[SAFE]: The instructions focus on mapping data models (e.g., donor records, gift transactions) and evaluating business logic (e.g., retention rate calculations, lapse risk scoring) which are consistent with the stated purpose of donor retention analysis.
[SAFE]: The 'Self-Evolution Telemetry' section contains a conditional check for a local project memory directory to log execution metadata. This is a standard observability pattern for autonomous agents and does not involve exfiltration of sensitive data to external domains.
[SAFE]: Input handling via $ARGUMENTS is used to scope the analysis (e.g., 'lapse risk scoring') and does not demonstrate susceptibility to command injection or prompt override techniques.

donor-retention