reset-context-contamination
PassAudited by ClawScan on May 14, 2026.
Overview
This is an instruction-only reset skill that coherently tells the agent how to reframe a conversation, with no code, credentials, install steps, or hidden data access.
This skill appears safe to install as an instruction-only aid. Use it when you want the agent to ignore prior drafts and rebuild the task from a clean brief, and be mindful that a summarized brief could be shared with a fresh subagent or new session.
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.
If used, the agent may disregard earlier reasoning or drafts and start from a distilled brief instead.
The skill intentionally tells the agent to stop relying on prior drafts in the current thread. That is central to the reset purpose, but it changes normal context use.
Void the earlier drafts entirely - no referencing, quoting, or patching them.
Use this skill only when you genuinely want the agent to reset its framing; restate any constraints you still want preserved.
Facts from the current thread could be summarized and reused in a new agent context.
The skill may involve passing a distilled version of the thread to another agent or session. No hidden channel or code is present, but the handoff can carry information from the conversation.
Hand the extracted brief to a fresh subagent, or recommend I start a new session.
Before delegating to a fresh subagent or new session, avoid including secrets or private details that are not needed for the reset.