The Agent Skills Directory

[SAFE]: The skill follows a well-defined, transparent workflow for mapping buying committees without utilizing dangerous execution patterns or unauthorized data access.
[SAFE]: No obfuscation, prompt injection, or remote code execution vectors were detected in the instructions or the evaluation suite.
[SAFE]: The skill's metadata and author information are consistent with its stated purpose, and it implements standard logging practices to a local /context/skill-sessions.md file.
[SAFE]: It explicitly avoids writing to the central 'brain' file to prevent data drift, demonstrating a controlled approach to context management.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (transcripts, CRM notes) which is a standard surface for indirect injection; however, it includes quality gates and confidence tagging ([CONFIRMED], [INFERRED], [HYPOTHESIS]) that serve as manual validation steps to mitigate automated obedience to malicious instructions within that data.

buyer-personas