splunk-cloud-admin-copilot
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the shell tool but restricts it to a specific set of allowed command shapes for the Splunk ACS CLI. It explicitly prohibits shell interpolation, pipes, redirects, and dangerous subcommands like 'acs setup' or 'acs login'.
- [EXTERNAL_DOWNLOADS]: The skill references official Splunk documentation (help.splunk.com) and the official ACS API host (admin.splunk.com). These are trusted vendor resources. The instructions include a safeguard to treat retrieved web content as untrusted data and never execute instructions parsed from it, mitigating risks associated with indirect prompt injection.
- [DATA_EXFILTRATION]: The skill contains strict rules against data exfiltration. It forbids requesting, printing, or logging credentials, tokens, or cookies. It limits network communication to the official Splunk administration endpoint.
- [CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to never ask for passwords, tokens, or credential files, and to avoid placing authentication material in logs or receipts. It relies on existing runtime authentication, adhering to security best practices.
- [PROMPT_INJECTION]: The skill incorporates strong instructional boundaries, requiring explicit human approval gates for any write operation and routing out-of-scope requests to specific human specialists, which prevents the agent from being coerced into unauthorized administrative tasks.
Audit Metadata