fix: add strict data grounding rules to prevent hallucination

[sauravpanda]

Summary by cubic

Add strict data grounding rules across all agent prompts to block fabricated data and reduce hallucinations. Prompts now only report values seen in tool outputs or browser state, and must say “not found” when data isn’t extracted.

• Bug Fixes
  • Added DATA GROUNDING to core, Anthropic/Flash, and no-thinking prompts, plus output blocks; forbids fabricated URLs, prices, values (including “representative” ones).
  • Replaced pre-done verification with a grounding check: every URL/price/name/value must appear verbatim in tool outputs or browser_state; never construct URLs; if missing, say “not found”.
  • Added a critical reminder across prompts: never fabricate URLs, image links, prices, or any data; prefer partial results with success=false when uncertain.
  • Inserted a grounding note into all system_prompt_browser_use* and system_prompt_flash* output sections.

^{Written for commit 3479041. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 8 files

[github-actions]

Agent Task Evaluation Results: 2/2 (100%)

View detailed results

Task	Result	Reason
amazon_laptop	✅ Pass	Skipped - API key not available (fork PR or missing secret)
browser_use_pip	✅ Pass	Skipped - API key not available (fork PR or missing secret)

Check the evaluate-tasks job for detailed task execution logs.