setup-skills
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various CLI tools including
gh(GitHub),az(Azure DevOps),git, andnpxto perform repository discovery, manage issue tracker data, and handle project configuration. - [EXTERNAL_DOWNLOADS]: Facilitates the installation of agent skills from external GitHub repositories such as
mattpocock/skills,upstash/context7, andhumanlayer/skills. These downloads are part of the skill's primary setup function and use commit SHA pinning to mitigate supply chain risks. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Reads local project data from
docs/agents/harness.md,skills-lock.json, and CI workflow files (.github/workflows/*.yml). - Boundary markers: Relies on an instructional process but lacks explicit markers to delimit untrusted project content from internal instructions.
- Capability inventory: Capable of writing setup scripts and configuration files, and executing package installation tools.
- Sanitization: Does not explicitly sanitize or escape data read from the repository before using it to generate scripts or configuration.
- [DYNAMIC_EXECUTION]: Generates and manages shell scripts (
local-gate.sh,cloud-ship-bootstrap.sh) that use theevalcommand to execute shell strings derived from placeholders and environment settings populated during the setup process. - [PRIVILEGE_ESCALATION]: The provided
cloud-ship-bootstrap.shtemplate includes commands to install system-level packages usingsudo apt-get install, which is a standard procedure for configuring CI environments.
Audit Metadata