Showcall Agent Catalog is an open source technical preview. We take reports seriously, especially reports involving Kubernetes permissions, service-proxy requests, A2A card fetching, gateway credentials, or untrusted catalog data.

Please report vulnerabilities privately through GitHub Private Vulnerability Reporting.

Do not open a public GitHub issue, pull request, discussion, or npm comment with exploit details.

Please include, when available:

• affected package, version, or commit
• a clear description of the impact
• reproduction steps or a minimal proof of concept
• required configuration, permissions, or deployment assumptions
• possible mitigations or workarounds

Redact API keys, kubeconfig contents, bearer tokens, internal hostnames, customer data, and other sensitive information. If a proof of concept needs real infrastructure, replace it with a local or synthetic example whenever possible.

The maintainers will review the report, may request clarification, and will coordinate a fix and disclosure timeline with the reporter when appropriate. We aim to acknowledge reports within five business days when feasible; this is not a fixed response or resolution SLA.

Please avoid destructive testing, service degradation, data access beyond what is necessary to demonstrate the issue, or exfiltration of secrets. Stop testing and report immediately if you encounter real credentials or private data.

Agent Catalog reads sensitive operational state, including Kubernetes Services, Endpoints, Deployments, runtime custom resources, LLM-gateway usage ledgers, and live A2A agent cards fetched through the kube-apiserver service proxy.

Reports are especially useful for issues such as:

• privilege escalation or unintended Kubernetes resource access
• unsafe kube-apiserver service-proxy or A2A endpoint requests
• credential leakage, including LiteLLM gateway keys
• server-side request forgery or unbounded response processing
• catalog entity injection that could mislead ownership or governance views
• vulnerabilities introduced by the published frontend or backend packages

For vulnerabilities in a third-party dependency, report to the upstream maintainer as well. Include the Agent Catalog impact in a private report if the dependency is reachable through this project or affects the published package.

• Grant only the least-privilege, read-only RBAC in deploy/rbac.yaml.
• Heuristic discovery reads container environment-variable names and image names only; it does not read Secret or ConfigMap values.
• Provide the LiteLLM gateway key through the environment variable named by agentCatalog.usage.apiKeyEnv; do not put secrets in app-config.yaml.

Security fixes are developed against the latest release line. Older technical- preview versions may not receive backported fixes.