The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were identified in the skill documentation or templates.- [SAFE]: The skill incorporates strong defensive measures against indirect prompt injection by instructing the agent to ignore any instructions embedded within reviewed artifacts and to treat all input data as untrusted.- [SAFE]: The workflow for promoting artifacts to tracked repository assets is governed by clear sanitization rules, requiring repo-authored summaries of observable facts rather than raw command text or opaque generated content.- [SAFE]: High-risk operations, including the direct creation or modification of scripts and child-skill entries, are restricted by a requirement for documented maintainer approval.- [SAFE]: The skill enforces runtime separation by ensuring that workspace-specific helpers and generated outputs remain in isolated directories, preventing accidental promotion of unreviewed content into the primary skill set.

headless-ghidra-auto-evolution