oodle-discovery
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill offers to install the
uvPython package manager using its official installation script fromastral.shif the tool is missing. This is a well-known, high-reputation service, and the download is part of the intended environment setup. - [COMMAND_EXECUTION]: Collector scripts (e.g.,
collectors/k8s/collect.py,collectors/gcp/collect.py) utilizesubprocess.runandsubprocess.Popento interface with system binaries includingkubectl,gcloud, andaws. These commands are hardcoded and used exclusively for read-only metadata discovery as described in the skill's core rules. - [PROMPT_INJECTION]: The skill ingests untrusted data from cluster configurations and cloud metadata (e.g., pod names, CRDs). It effectively mitigates indirect prompt injection risks by instructing the agent to treat this data as unverified and by performing HTML escaping in
report/template.pybefore rendering the final report.
Audit Metadata