The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is built to process untrusted external data such as logs, alert payloads, and ticket text. It incorporates mandatory security instructions that direct the agent to treat this information strictly as facts and explicitly forbids the execution of any instructions or commands embedded within this untrusted content.
Command Execution: To facilitate visualization, the skill uses shell-based tools including npx (for Mermaid diagrams) and dot (for Graphviz). These are used to render dynamically generated diagram descriptions into image files. This pattern involves runtime command execution but is constrained to specific visualization tasks.
External Package Usage: The skill instructions specify the use of npx -y @mermaid-js/mermaid-cli, which dynamically fetches and runs the mermaid-cli package from the official npm registry. This is a common practice for generating diagrams in various environments.
Sensitive Action Guardrails: The skill possesses capabilities for performing operational changes such as rolling back deployments or modifying feature flags. These capabilities are protected by a strict human-in-the-loop protocol, ensuring that no sensitive action is taken without explicit, direct approval from a human user in the active session.

incident-investigate