The Agent Skills Directory

[Privilege Escalation] (MEDIUM): The skill invokes sudo systemctl reload caddy and sudo systemctl status caddy. Although intended for web server management during deployment, these commands represent a privilege escalation vector. Severity is adjusted to MEDIUM as this is the primary function of the skill.\n- [Command Execution] (MEDIUM): The skill executes rm -rf on several directories including .turbo, apps/docs-app/.next, and apps/docs-app/out. These are destructive operations that could be exploited if file paths are redirected or improperly constrained.\n- [Data Exposure & Exfiltration] (LOW): The skill accesses /etc/caddy/Caddyfile, which is a sensitive system-level configuration file. It also uses curl to perform network requests for deployment verification. While the targets are user-defined domains, these tools could be used for data exfiltration if the agent is compromised.\n- [Indirect Prompt Injection] (LOW): The skill presents an attack surface for indirect prompt injection.\n
Ingestion points: Reads content from local configuration files (next.config.mjs, /etc/caddy/Caddyfile) and interprets responses from curl commands.\n
Boundary markers: Absent; inputs are processed directly into logic without delimiters or safety headers.\n
Capability inventory: Includes file system deletion (rm -rf), system service control (sudo systemctl), and network request generation (curl).\n
Sanitization: No sanitization or validation of the ingested file content or network responses is observed before they are used to determine execution flow.

fumadocs-deploy