The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill processes untrusted source files, providing an attack surface where embedded instructions could influence the agent. Ingestion points: Glob, Grep, and Read tools applied to code. Boundary markers: Absent. Capability inventory: Bash, Write, and Rube MCP (Slack/Jira). Sanitization: No explicit filtering of processed code content.
[Command Execution] (LOW): Utilizes the Bash tool for external analysis and troubleshooting. This is consistent with the skill's purpose but represents a significant capability.
[Data Exfiltration] (LOW): Capability to send findings to Slack or Jira via mcp__rube__RUBE_MULTI_EXECUTE_TOOL enables data transfer to external domains, which requires monitoring.

sc-analyze