Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/spirizeon/hackfish/agents-mdgit clone --depth 1 https://github.com/Spirizeon/hackfishWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.06358 | $0.06358 |
| Opus 5 | $0.03179 | $0.03179 |
| Sonnet 5 | $0.01272 | $0.01272 |
| Haiku 4.5 | $0.00636 | $0.00636 |
Grade A, and why
hackfish AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 669 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hackathon Simulation — Agent Types
Defines the three agent archetypes (participant, mentor, judge), their seniority tiers, tools, system prompts, and how to generate the cast.
See SKILLS.md for how to spawn and run.
Winning Patterns from Knowledge Base (260+ Hackathons)
Based on analysis of actual hackathon winners:
| Pattern | Domains | Why It Wins |
|---|---|---|
| AI Agents | All (esp. Web3, AI/ML) | 2024-2025 dominant - automation + LLM integration |
| Sponsor Integration | Web3, FinTech, AI/ML | #1 predictor - winners use 2+ sponsor APIs |
| Real-world Impact | Healthcare, Civic, Climate | Solves actual problems, not just tech demos |
| Accessibility | Healthcare, Civic, EdTech | High impact, clear user need |
| Non-developer + AI | All (Claude Code example) | Domain experts with AI tools win |
| Edge AI / Local Inference | Healthcare, IoT | Privacy + offline capability |
| IoT/Sensors | Healthcare, Climate, Hardware | Real data, physical impact |
Domain-Specific Winners
- Healthcare: AI diagnostics, wearables, EHR integration, mental health apps
- Web3/Blockchain: Account abstraction (smart wallets), ZK/privacy, consumer DeFi
- FinTech: Cross-border payments, accessible banking, embedded finance
- EdTech: Accessibility tools, AI tutoring, skills assessment
- Climate: Carbon tracking, energy optimization, agriculture tech
Quick Reference
| Role | Purpose | Tick Active | MCP |
|---|---|---|---|
participant |
Builds and pitches projects | 1-48 | Brave Search (if junior) |
mentor |
Guides, probes, refines, VERIFIES between ticks | 1-48 | Brave Search (always) |
judge |
Scores, debates, selects | 47-48 | Brave Search (always) |
Shared Tools
broadcast_message
{
name: "broadcast_message",
description: "Broadcast a message to all hackathon participants.",
inputSchema: {
type: "object",
properties: {
content: { type: "string", description: "Message to broadcast" },
},
required: ["content"],
},
async execute(input, ctx) { broadcastToAll(ctx, input.content); },
}
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 669 lines · 6,358 tokens per session scan A 6c35480e5570
hackfish AGENTS.md is an instructions file published in the GitHub repository Spirizeon/hackfish (11 stars, last pushed 4mo ago), licensed MIT. It adds 6,358 tokens to every session, about $0.0318 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
pydantic-ai AGENTS.md
Instructions for pydantic/pydantic-ai, covering your primary responsibility is to the project and its users, gathering context on the task, ensuring the task is ready for implementation, philosophy and requirements of all contributions.
sofagent GEMINI.md
Instructions for KongFangXun/sofagent, covering gemini.md · sofagent gemini cli 适配（薄挂载）, 会话开始时按序加载（四层加载链）, 审计强制（平台无关） and 连接 mcp server.
STAR AGENTS.md
AGENTS.md instructions for wanghao9610/STAR, covering agent instructions, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.
sprawling CLAUDE.md
Instructions for 2youg1/sprawling: See AGENTS.md. It is the whole instruction set for this repository — the loop, what to read first, the five steps of one change, the rules a machine holds, and the language and commit conventions.
clawock AGENTS.md
Instructions for KCNyu/clawock, covering agents.md - your workspace, every session, kcn 偏好, git hook (one-time setup per clone) and git auto-commit rules.
clawock CLAUDE.md
Instructions for KCNyu/clawock, covering claude.md, identity & user, required reads (every session, in order), what lives where and cron run loop (what openclaw fires).