The Agent Skills Directory

[SAFE]: The skill logic is purely instructional, focusing on conducting a persistent and constructive business interview. It contains no malicious commands, hidden scripts, or unauthorized network operations.
[DATA_EXFILTRATION]: The skill instructions explicitly forbid the agent from performing external actions, sending data, or connecting to systems without direct user confirmation, preventing unauthorized data exfiltration.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest data from local project files and user responses. However, this risk is mitigated by the structured interview workflow and the requirement for user approval for all material actions.
Ingestion points: Reads .business/context.md, .business/metrics.md, and user-supplied evidence as instructed in SKILL.md.
Boundary markers: No explicit markers defined for ingested content.
Capability inventory: Access to read project files and write to the .business/ directory.
Sanitization: No explicit sanitization or filtering of external context content is defined.

grilling