The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted participant responses during the facilitation and review stages, which creates a potential surface for indirect prompt injection.
Ingestion points: Interviewee responses are collected in 'facilitate' mode and analyzed in 'review' mode as part of the evidence-led process.
Boundary markers: The instructions do not define specific delimiters (such as XML tags) or explicit 'ignore embedded instructions' warnings for isolating the participant-provided text from the skill's operational instructions.
Capability inventory: The skill is restricted to drafting observations and findings; it does not request high-risk tools for network operations, file system modification, or command execution.
Sanitization: There is no mention of automated sanitization or filtering, though the skill relies on mandatory human review of all AI-generated drafts.
[SAFE]: The skill implements strict ethical and operational guardrails, including explicit 'Stop conditions' that halt the process if safety criteria are not met, such as missing accommodations or requests for prohibited inferences regarding health, personality, or protected traits.
[SAFE]: Accountability is strictly maintained by requiring that all outcomes and decisions be made by a named human evaluator, explicitly forbidding the AI from ranking candidates, sending rejections, or acting as an autonomous decision-maker.

async-interview