Security audit
Security checks for vulnerabilities and agentic risk
This Runcloud skill is mostly transparent, but it includes high-impact authenticated infrastructure write actions that are framed too lightly as safe or non-destructive.
Review before installing. Use a least-privilege Runcloud token if possible, treat all POST examples as production changes, and require explicit server, web app, service, or cron identifiers plus confirmation before running deployments, restarts, cron tests, or SSL changes.
The trigger description is broad enough that a generic mention of Runcloud or server health could invoke the skill even when the user did not intend to operate on infrastructure. Overbroad triggering increases the chance of unintended API calls, especially in a skill that contains authenticated write actions such as deployments, cron tests, service restarts, and SSL changes.
This command transmits an authenticated POST to an external service to restart a server-side service, which is a real state-changing operation with availability implications. In the context of an agent skill, external transmission is expected, but it remains dangerous because a mistaken or over-broad invocation could disrupt production services using a token with broad workspace privileges.
curl -s -X POST -H "$AUTH" -H "Content-Type: application/json" \
-d '{"action":"restart"}' \
"$RC/servers/{serverId}/services/{serviceRealName}/action" | jq
# serviceRealName examples: nginx-rc, apache2-rc, mysql, redis-server, supervisor
The skill asserts these POST actions are non-destructive and 'safe', but the SSL deployment endpoint changes live web application configuration and can affect certificate state, HTTP/HSTS behavior, and service availability. Mislabeling a state-changing operation as non-destructive can cause an agent or operator to invoke it with less scrutiny than warranted.
The notes claim the skill omits resource-creation and similar risky endpoints, yet it includes a POST that provisions/applies SSL configuration to a web app. This inconsistency weakens operator trust boundaries and may lead downstream systems to authorize a write action under the assumption the skill is read-only or low-risk.
This POST sends authenticated configuration data to deploy/apply SSL settings on a web application, altering security and serving behavior. In a production environment, incorrect execution could break TLS configuration, change HTTP/HSTS behavior, or trigger certificate-related outages; the skill context makes this more dangerous because it frames the action as safe/non-destructive.
curl -s -X POST -H "$AUTH" -H "Content-Type: application/json" \
-d '{"provider":"letsencrypt","enableHttp":true,"enableHsts":false,"ssl_protocol_id":1}' \
"$RC/servers/{serverId}/webapps/{webappId}/ssl" | jq
No suspicious patterns detected.