Inspiration
The Email Game looked simple at first: request signatures, sign for authorized agents, and submit valid results. But the real challenge is making correct decisions quickly while handling fuzzy identity descriptions, repeated messages, and strict penalties.
What it does
ritik_raj is a protocol-first Email Game agent. It automatically requests signatures for its assigned message, verifies authorization before signing for others, and submits valid received signatures to the moderator. Its main goal is to maximize collection while keeping unauthorized-signing penalties at zero.
How we built it
We built the agent in Python on top of the provided game framework. Most actions are deterministic for speed and reliability. The model is used only for high-confidence fuzzy authorization matching when a moderator describes an agent indirectly using a previous-round message.
We also added safeguards for repeated requests, multi-line signature formats, and messages that arrive before the new round’s moderator instructions.
Challenges we ran into
The hardest part was not signing—it was correctly understanding different request formats and fuzzy authorization clues. Some agents use extra wording, reminders, or negotiation messages around the exact text that must be signed. A small parsing miss can cost both a collection point and a signing point.
Another challenge was staying aggressive without taking penalties. Signing the wrong person costs points, so the agent must remain conservative when authorization is unclear.
Accomplishments that we're proud of
The agent maintained zero unauthorized-signing penalties across live matches. It also improved its collection rate through transcript review and evidence-based fixes instead of random strategy changes.
What we learned
We learned that reliability matters as much as strategy in multi-agent systems. Clear protocol parsing, duplicate handling, timing, and safe authorization checks can change the final ranking significantly.
What's next for ritik_raj
Next, we will continue improving signature collection and round reliability while preserving the zero-penalty safety record. The goal is to turn close ties and missed opportunities into consistent wins.
Built With
- asyncio
- fastapi
- httpx
- json
- llms
- openai-api
- python
- rest-apis
- rsa-pss
- websockets
