product-roadmap
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via $ARGUMENTS and project-specific artifacts to author roadmap documentation and scope briefs. While the skill does not use automated content sanitization, it implements multiple 'Decision Gates' that require explicit human approval via the 'ask user question tool' before artifacts are finalized or implementation planning begins.\n
- Ingestion points: The skill processes user-supplied input through the $ARGUMENTS variable in SKILL.md and reads project files such as docs/project-config.json and .claude/skills/shared/product-roadmap-contract.md.\n
- Boundary markers: The workflow includes an 'Explicit Invocation Gate' in Phase 0 and a mandatory 'Decision Gate' in Phase 3 to verify user intent and approve generated content.\n
- Capability inventory: The skill is capable of writing markdown files to the local file system (e.g., docs/product-roadmap.md and scope-brief.md) and invoking downstream agent skills including $scenario, $brainstorm, and $plan.\n
- Sanitization: No formal programmatic sanitization is defined; the skill relies entirely on human-in-the-loop validation to identify and correct misinterpretations or malicious instructions embedded in the source data.
Audit Metadata