The Agent Skills Directory

[SAFE]: The skill contains no malicious code, obfuscation, or persistence mechanisms. It is designed with a strong focus on safety and non-execution.- [INDIRECT_PROMPT_INJECTION]: While the skill ingests potentially untrusted external data such as security advisories and logs, it employs significant mitigations. It explicitly instructs the agent to treat such data as 'untrusted evidence' and requires the use of an evidence ledger to track the provenance of all facts (e.g., 'user-provided' vs 'document-backed').- [DATA_EXFILTRATION]: The skill utilizes the web tool for retrieving documentation. However, it explicitly prohibits the request or collection of sensitive data such as credentials, private support content, or raw customer data. The network operations target official Splunk domains, which are recognized as legitimate vendor resources for this skill.- [COMMAND_EXECUTION]: The skill is strictly limited to an advisory role. The instructions repeatedly state that it must not execute, schedule, or approve any changes, nor should it perform active incident remediation or change production configurations.

upgrade-and-vulnerability-readiness-advisor