product-spec
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill focuses on documentation and product decision-making, with explicit prohibitions against source code modification.
- Evidence: The 'Hard Rules' section in SKILL.md forbids editing source, staging, committing, or running implementation.
- [COMMAND_EXECUTION]: The skill performs read-only repository status checks to gather context.
- Evidence: Workflow step 1 in SKILL.md requires running
git status --shortto inspect affected consumers and existing repository facts. - [DATA_EXFILTRATION]: The skill includes a capability for external research that involves sending data to another model.
- Evidence: The Output Contract in SKILL.md describes handing fixed questions to the
ask-chatgpttool. - Mitigation: The skill mandates that this never happens implicitly and requires a fixed question or basis authorized by the user.
- [PROMPT_INJECTION]: The skill processes untrusted user requirements and repository data, creating a potential surface for indirect prompt injection.
- Evidence: The workflow involves reading 'requested scope' and 'existing product facts' to generate specifications.
- Mitigation: The skill implements a material statement classification system (Confirmed, Assumption, Open Question, Rejected, Deferred) and follows strict templates in references/template.md to ensure boundaries are maintained.
Audit Metadata