Rationality
PassAudited by ClawScan on May 10, 2026.
Overview
This is a documentation-only reasoning framework with no code, install steps, credentials, or network behavior, though it does suggest persistent memory updates and supervised local debugging/revert actions.
This skill appears safe as an instruction-only reasoning aid. Before using it, decide whether you want the agent to store corrections in memory, and keep approval enabled for any local commands that could inspect or modify your workspace.
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 the host agent supports memory, corrections or critiques could influence future tasks beyond the current conversation.
This instructs the agent to persist corrected reasoning for reuse. That supports the self-correction purpose, but the artifacts do not define what may be stored, retention limits, or user approval.
When an error is corrected, record the "Refutation" in `memory/`.
Allow memory writes only with clear user consent, store minimal non-sensitive summaries, and provide a way to review or delete them.
An agent applying this guidance in a coding workspace could revert local work if tool execution is allowed.
The skill suggests local state-changing repair actions. This is coherent with error correction, but such commands can affect user files if executed without confirmation.
Revert to the last known good state (e.g., `git reset`).
Require explicit user confirmation and a backup/status check before running destructive or state-changing commands such as git reset.