create-issue-interaction-ui
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The instructions include shell commands for viewing commit details (
git show) and executing the project's test suite usingpnpmandvitest. These commands are standard for a software development workflow within a local repository. - [INDIRECT_PROMPT_INJECTION]: The skill provides a blueprint for creating interaction cards (e.g., forms, confirmations) that ingest external data, which is an attack surface for indirect prompt injection.
- Ingestion points: Data is received via user interaction payloads and result interfaces defined in
packages/shared/src/types/issue.ts. - Boundary markers: Prompt instructions for agents are managed in
skills/paperclip/SKILL.mdto define how the agent should interpret different interaction kinds. - Capability inventory: The server-side interaction service handles the lifecycle of these requests, including state transitions and activity logging.
- Sanitization: The skill explicitly requires the implementation of Zod schemas in
packages/shared/src/validators/issue.tsto validate and sanitize all incoming payloads.
Audit Metadata