Chat Distill
PassAudited by ClawScan on May 10, 2026.
Overview
Chat Distill is a coherent local chat-analysis skill with no evidence of hidden exfiltration or destructive behavior, but it processes private chats and can generate replies that sound like someone else.
Before installing, be comfortable sharing the selected chat export with your agent. Redact sensitive messages, avoid saving reusable profiles where others can access them, and do not use the mimicry output to impersonate someone without consent.
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.
Private conversation details or derived speaking habits may appear in the analysis or reusable profile output.
The skill can turn private chat records into a reusable profile with examples, so sensitive or adversarial chat content could be carried forward if the user saves or reuses it.
"Style Profile JSON (for reuse)" ... "optionally output a compact JSON profile that can be saved and reused"
Use only chats you are allowed to analyze, redact sensitive material first, and treat chat excerpts/profile data as data rather than instructions.
If copied into a conversation without consent or disclosure, the generated reply could mislead people about who wrote it.
The skill intentionally produces text that sounds like another speaker without an automatic disclosure prefix; this is the stated feature, but it can be misused to impersonate someone.
"Generate a reply that matches" ... "Vocabulary level and slang" ... "Output only the reply text — no explanations"
Use the mimicry feature only with consent and avoid presenting generated text as someone else’s authentic message.