template-builder
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of GitHub templates by ingesting user responses and incorporating them into structured YAML output, creating a vector for potential indirect prompt injection attacks. \n
- Ingestion points: User input is collected through interactive questions about template metadata (e.g., name, description, labels) and field attributes (e.g., labels, help text) as described in Phase 1 and Phase 2 of SKILL.md. \n
- Boundary markers: There are no explicit instructions for the agent to use specific delimiters or boundary markers to isolate user-provided strings from the template structure. \n
- Capability inventory: Behavioral rules in SKILL.md imply the agent has capabilities to interact with the file system, such as checking for existing templates and potentially writing files to the .github/ISSUE_TEMPLATE/ directory. \n
- Sanitization: The agent is instructed to format field IDs but lacks instructions to sanitize or escape other user-provided strings, such as descriptions or markdown text, which could contain malicious instructions.
Audit Metadata