security-threat-model
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from local source code repositories to generate security summaries and threat models.
- Ingestion points: The workflow in
SKILL.mdand the templates inreferences/prompt-template.mddirect the agent to read files and directory structures from a user-provided repository path. - Boundary markers: While the
prompt-template.mddefines a specific persona and task, it lacks robust boundary markers or "ignore instructions" delimiters to prevent the agent from following malicious prompts embedded within code comments or documentation in the analyzed repository. - Capability inventory: The skill utilizes file-reading capabilities (including recommended use of
rg/ Ripgrep) and writes a final report to the local filesystem (<repo-or-dir-name>-threat-model.md). - Sanitization: The skill includes a proactive safety rule in
references/prompt-template.mdto redact any secrets, tokens, or passwords encountered during analysis.
Audit Metadata