sdlc
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it processes untrusted external data that is then used to influence subsequent agent actions.
- Ingestion points: As defined in
references/phase-1-intent.md, the skill captures the user's initial development request "verbatim" and stores it within theintent.mdartifact. - Boundary markers: The artifact templates (e.g.,
assets/intent-template.md) lack explicit delimiters or "ignore embedded instructions" warnings to prevent the agent from following malicious commands hidden within user-provided text or repository evidence. - Capability inventory: The skill is capable of writing multiple Markdown and HTML files to the local filesystem and invoking platform tools like
Scout(for repository analysis) andOracle(for architectural decisions). - Sanitization: There are no instructions provided for filtering, escaping, or validating the user-authored requests or repository content before they are interpolated into the SDLC artifacts.
Audit Metadata