build-in-public

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill uses platform-integrated tools (linear, gh, git, devsql) to access project metadata and development logs. Access is restricted to the user's authenticated scope.
  • [SAFE]: A manual review gate is enforced; the skill is instructed never to post updates automatically, leaving final validation and distribution to the user.
  • [SAFE]: Indirect Prompt Injection Risk Assessment:
  • Ingestion points: The skill processes external content from Linear issues, GitHub pull requests, and Git commit logs (SKILL.md).
  • Boundary markers: While no explicit data delimiters are used, the output format is limited to a text draft for the user.
  • Capability inventory: The skill can only query project status and local session history; it lacks capabilities for persistent storage, file modification, or outbound network calls to arbitrary domains.
  • Sanitization: Data is transformed through summarization and synthesis into a specific update format, which provides an inherent layer of abstraction from source inputs.
  • Risk evaluation: The overall risk is low because the agent's actions are confined to text generation for human approval.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 11:38 AM
Security Audit — agent-trust-hub — build-in-public