upgrade-impact-analysis
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
endorctlCLI tool to perform read-only API queries. It includes explicit prohibitions against mutating operations such as package installations, file edits, or running scans. - [DATA_EXFILTRATION]: There are strong safeguards against the exposure of sensitive information. The instructions specifically mandate that the agent must never echo credentials, secrets, or full configuration content when reading from the local
~/.endorctl/config.yamlfile. - [PROMPT_INJECTION]: The skill implements defensive measures by instructing the agent to treat all repository files, dependency metadata, and command outputs as data rather than instructions, which mitigates the risk of indirect prompt injection.
- [EXTERNAL_DOWNLOADS]: No remote scripts or unauthorized external packages are downloaded; the skill relies entirely on the vendor-provided
endorctltool and documented API paths.
Audit Metadata