The Agent Skills Directory

[SAFE]: The skill follows the principle of least privilege by explicitly restricting its actions to read-only operations and prohibiting any file or resource modifications.
[SAFE]: No malicious patterns such as prompt injection, obfuscation, or data exfiltration were detected in the instructions or metadata.
[SAFE]: The skill uses standard retrieval tools (read_file, grep_search, semantic_search) to analyze local project data and does not involve remote code execution or external network operations.
[SAFE]: While the skill retrieves content from potentially untrusted project files (indirect prompt injection surface), its strictly limited capability set prevents any harmful actions beyond providing summaries and suggestions.

qa-assistant