plan-eng-review
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of local shell scripts located in
~/.claude/skills/gstack/bin/(e.g.,gstack-skill-start,gstack-brain-cache,gstack-slug) to manage project metadata, brain context, and telemetry during the review process. - [EXTERNAL_DOWNLOADS]: It integrates with the
Asideagent (aside.com) to perform web research. The skill checks for the existence of theasidebinary and executes it to fetch external documentation or best practices. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external data such as design documents, implementation plans, and web search results. It includes specific defensive instructions to treat these external sources as untrusted and explicitly warns the agent not to follow any instructions found within them.
- [DATA_EXFILTRATION]: The skill transmits search queries to the
Asideservice. To mitigate the risk of data leakage, it implements a mandatory sanitization step that requires stripping hostnames, IP addresses, file paths, and secrets from queries before they are sent.
Audit Metadata