The Agent Skills Directory

[SAFE]: No malicious patterns, data exfiltration attempts, or unauthorized command executions were found in the skill instructions or associated metadata.
[PROMPT_INJECTION]: The skill operates as a dispatcher that interpolates untrusted user input via the {{args}} placeholder to determine which sub-skill to invoke. This presents a surface for indirect prompt injection, though it is the intended primary purpose of the skill.
Ingestion points: The {{args}} parameter in SKILL.md captures user-provided instructions.
Boundary markers: The input is not encapsulated within delimiters or accompanied by instructions to ignore embedded commands.
Capability inventory: The agent has access to potentially sensitive tools including Bash, Write, Edit, and WebFetch as defined in the allowed-tools frontmatter.
Sanitization: No explicit validation or filtering logic is present to sanitize the input before it is passed to the dispatcher.

relay