The Agent Skills Directory

[SAFE]: The skill is comprised of markdown playbooks, CSV routing configuration, and output templates. No security issues were detected after a thorough review of the instructions and supporting scripts.
[PROMPT_INJECTION]: The skill processes user-supplied data such as interview transcripts and notes, which creates an inherent surface for indirect prompt injection. However, the risk is negligible as the skill lacks high-privilege capabilities such as network access or shell execution.
Ingestion points: SKILL.md (intents: critique-questions, conduct, synthesize) where user data is ingested for analysis.
Boundary markers: The instructions explicitly mandate the separation of behavioral evidence from user claims, providing a logical framework for data processing.
Capability inventory: The agent is limited to generating markdown files based on provided templates (templates/interview-plan.md, templates/interview-snapshot.md) and modifying text inline.
Sanitization: No explicit automated sanitization of external content is described, but the instructional logic focuses on evidence-based extraction which limits the impact of malicious content in the input.

customer-interviewing