Elicitation - how to talk with humans and ask them questions?
ReviewAudited by ClawScan on May 10, 2026.
Overview
This instruction-only skill has no code or install risk, but it is designed to elicit and infer sensitive psychological information through ordinary conversation, so it needs strict consent and privacy limits.
Before installing, decide whether you want an agent to conduct psychologically sensitive interviews. Use it only with clear participant consent, avoid diagnostic or therapy-like claims unless qualified, keep sensitive topics optional, and do not store or share profiles without explicit permission.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A person could reveal intimate memories, beliefs, vulnerabilities, or identity information without understanding that the agent is profiling them.
The skill explicitly teaches rapport-based psychological profiling and disclosure elicitation, but the provided instructions do not make explicit informed consent or disclosure to the person being profiled a hard requirement.
build psychological profiles through gradual disclosure ... elicit authentic self-disclosure without interrogation
Use only in consent-based interview or personal-discovery contexts, clearly disclose the profiling purpose, let people skip or stop, and avoid using the techniques for covert persuasion or manipulation.
Sensitive or mistaken psychological inferences could influence later interactions if they are retained or reused.
The skill correctly warns against over-trusting language inference, but it also contemplates aggregating psychological observations across conversations; if saved in memory or notes, those observations become sensitive profile data.
Cross-validation is essential. Never rely on language alone. Triangulate with: ... Multiple conversations ... Different contexts.
Do not store psychological profiles by default; require explicit consent for memory use, label inferences as tentative, and provide a way to review or delete retained notes.
