The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external lecture files and project rules located in .claude/rules/. While these inputs represent a potential attack surface for indirect prompt injection, the skill is explicitly restricted to read-only tools (Read, Grep, Glob) and lacks network access. This configuration prevents any malicious instructions within the processed data from performing unauthorized actions, exfiltrating data, or modifying the filesystem.
[SAFE]: The skill's behavior is consistent with its stated purpose of providing adversarial pedagogical reviews. No evidence of obfuscation, credential harvesting, privilege escalation, or remote code execution was found. The use of YAML frontmatter to restrict tools to a read-only scope follows security best practices for AI agent skills.

devils-advocate