complete-umbrella
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from issue bodies, CI output, and sub-agent results. It mitigates this risk through clear instructional boundaries in
SKILL.mdandphase-common.md, mandating that this data must never select commands or supply shell text, and using isolated temporary directories for all phase-specific data handoffs. - [EXTERNAL_DOWNLOADS]: The skill interacts with GitHub and Git remotes to fetch issue data and synchronize code. These operations are performed using standard tools (gh, git) via a controlled vendor script (
larch.sh) and are restricted to the authorized repository context. - [COMMAND_EXECUTION]: Shell commands are consistently routed through the vendor-provided
${CLAUDE_PLUGIN_ROOT}/scripts/larch.shscript, which manages background jobs, repository state, and implementation lifecycle steps in a controlled manner.
Audit Metadata