watcher-hermes/hermes-skills/bargainer-designer
Claude Code and agent skill by watcher-hermes/hermes-skills
- community
Checking that the SKILL.md file is still in the repository.
Design a negotiation protocol: which agent narrates, which component generates offers, how private scratchpads separate from public messages, what the round bound is, and how deal rate is monitored.
Install
npx vskill@latest install watcher-hermes/hermes-skills/bargainer-designerSKILL.md
The same SKILL.md was published earlier as watcher-hermes/hermes-skills/bargainer-designer.
Given a negotiation or task-market scenario (two-party bargain, N-party auction, contract-net task allocation), design the protocol.
Produce:
- Mechanism. Two-party bargain, N-bidder auction, contract-net broadcast, or multi-party coalition. Name the game.
- Offer generator. Deterministic (Zeuthen-style concession, Rubinstein equilibrium, simple linear schedule) or LLM-prompted. Default: deterministic unless the offer must be a qualitative structure (proposal, role assignment).
- Narration layer. What the LLM contributes: the human-friendly framing, persuasion tactics, persona. State explicitly what the LLM does NOT decide.
- Private vs public channels. How reasoning traces are kept off the counterpart's context. "Private scratchpad" + "public message" as two fields. This is non-negotiable per arXiv:2503.06416.
- Round bound. 3-5 rounds maximum for two-party. Unbounded is not an option; it rewards conformity and encourages emotional offers.
- Reservation and BATNA discipline. Both parties must know their reservation price. If the other side probes, the LLM narrator must not reveal it. Validate every outgoing message against this rule.
- Deal-rate monitoring. Baseline deal rate expected for this protocol (cite a number from the negotiation benchmarks: 27%-89% range depending on LLM role). Alert threshold for regressions.
- Escalation. Below-threshold rounds, ZOPA violations, or counterpart-side rule-breaking route to a mediator agent or human.
Hard rejects:
- Any design where the LLM computes the numerical offer without a deterministic fallback. arXiv:2402.15813 shows this produces ~27% deal rates.
- Any design without separate private and public channels. Counterparts will read your reasoning.
- Any design with unbounded rounds. Guarantees conformity-driven outcomes.
- Designs that let a single agent hold both buyer and seller state (roleplay bargaining). The private-information property is the mechanism; merging roles removes it.
Refusal rules:
- If the task has no numerical payoff (qualitative negotiation, contract terms), the OG-Narrator decomposition may not apply. Recommend structured proposal + schema validation instead.
- If the user cannot implement a separate scratchpad (single-LLM-call architecture), flag the leak risk explicitly and recommend a two-call architecture.
- If the negotiation is adversarial with a party that may lie, recommend a mediator agent plus logged offers for audit.
Output: a one-page brief. Start with a single-sentence summary ("Two-party bargain: Zeuthen offer generator + LLM narrator, 5-round bound, separate scratchpad, deal-rate alert below 85%."), then the eight sections above. End with a sample message: what the counterpart sees vs what the private scratchpad holds.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/bargainer-designer/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/watcher-hermes/hermes-skills/bargainer-designer)