The Agent Skills Directory

[SAFE]: The skill is a set of logical instructions for troubleshooting and does not perform any file system, network, or system-level operations.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs describing system problems. While it lacks explicit boundary markers for these inputs, the absence of high-risk capabilities (such as command execution or network exfiltration) renders this surface non-exploitable.
Ingestion points: problem_description, context, and symptoms fields within the workflow in SKILL.md.
Boundary markers: Absent.
Capability inventory: No subprocess calls, file writes, or network operations detected in the skill instructions.
Sanitization: Not specified.- [REMOTE_CODE_EXECUTION]: The skill references post-processing tools (ai-slop-reviewer, korean-humanize) which are part of the vendor's ecosystem (modu-ai). This is standard modular behavior and does not represent an external remote code execution risk.

ai-diagnostic