autoresearch

SUSPICIOUS. The skill is broadly capable and intentionally autonomous, with local shell-command execution and multi-step code modification that make it medium risk for misuse or unsafe operation. However, the provided text is internally aligned with its stated purpose, uses same-repo/local components, includes explicit approval gates for deploy/push, and shows no clear credential theft, hidden exfiltration, or malicious pre-execution behavior.

No explicit malicious code (e.g., embedded backdoor, credential theft, exfiltration routines, or obfuscated payload) is shown in this fragment. However, the workflow is intrinsically high-risk because it is designed to execute arbitrary user-supplied shell commands (Verify/Guard) and optionally chain into downstream commands while repeatedly mutating git history. If an attacker can influence the command arguments or chain targets, this template could be used for host compromise, sabotage, or data leakage via command side effects and/or metric/log outputs. Treat as a command-execution orchestration mechanism requiring strict allowlisting/sandboxing and least-privilege controls.