gsp-project-research

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill explicitly pre-fetches competitor/reference sites with WebFetch (Step 1.75) and instructs the spawned researcher to use WebSearch and incorporate pre-fetched third-party content into its prompt and research outputs (Step 2 and methodology/gsp-project-researcher.md), which means untrusted, public web content will be read and can materially influence agent actions and recommendations.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly calls WebFetch at runtime to pre-fetch "competitor URLs or reference sites" referenced in BRIEF.md or {PROJECT_PATH}/references/ and then inlines that fetched content into the spawned researcher agent's prompt, so any external URL listed there (i.e., competitor/reference site URLs fetched via WebFetch) can directly control the agent's instructions.