harbor
AGENTS.md6.7k starsChanged 24 days agoApache-2.0View on GitHub
This directory contains the Next.js landing page deployed at https://harborframework.com. Documentation is maintained in ../docs-mintlify and deployed at https://docs.harborframework.com. Use Bun for this application: - src/app/(home)/ owns the homepage and its components. - src/app/layout.tsx owns the root layout, metadata, fonts, and providers. - src/app/global.css contains Tailwind CSS and shared Fumadocs styling. - src/lib/layout.shared.tsx owns navigation options. - src/components/ contains shared UI components. - public/ contains website assets. - Use strict TypeScript and the @/* alias for src/*. - Use cn()…
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 9
Use Bun for this application:
- Very specific: line 27
Use strict TypeScript and the @/* alias for src/*.
- Very specific: line 28
Use cn() from src/lib/utils.ts to combine conditional classes.
- Distinctive: line 31
Search is disabled because documentation search belongs to Mintlify.
- Very specific: line 32
Do not edit generated .next/, node_modules/, or next-env.d.ts files.
- Very specific: line 36
Keep the permanent redirects in next.config.mjs:
- Very specific: line 42
Keep specific legacy redirects before the /docs/:path* fallback.
≈ “Common” is an estimate: at least 1 in 5 of the 36,368 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Harbor website
This directory contains the Next.js landing page deployed at
https://harborframework.com. Documentation is maintained in ../docs-mintlify
and deployed at https://docs.harborframework.com.
Development
Use Bun for this application:
bun install
bun dev
bun run build
bun start
bun lintStructure and conventions
src/app/(home)/owns the homepage and its components.src/app/layout.tsxowns the root layout, metadata, fonts, and providers.src/app/global.csscontains Tailwind CSS and shared Fumadocs styling.src/lib/layout.shared.tsxowns navigation options.src/components/contains shared UI components.public/contains website assets.- Use strict TypeScript and the
@/*alias forsrc/*. - Use
cn()fromsrc/lib/utils.tsto combine conditional classes. - Preserve the homepage's Fumadocs UI and core dependencies; they provide layout, theme support, styling, and code blocks. Fumadocs MDX is no longer used.
- Search is disabled because documentation search belongs to Mintlify.
- Do not edit generated
.next/,node_modules/, ornext-env.d.tsfiles.
Redirects
Keep the permanent redirects in next.config.mjs:
- Known
/docs/pages and historical aliases go directly to their corresponding Mintlify pages. Guides without a current equivalent go to their final pre-removal revision on GitHub. /docs/:path*sends unknown paths to the Mintlify documentation homepage.- Keep specific legacy redirects before the
/docs/:path*fallback. /news/:path*goes to the corresponding Mintlify news path./llms-full.txtgoes to Mintlify's text export./registryand/registry/:path*go to Harbor Hub.
The docs/ directory remains the website's deployment root. Add documentation
pages to docs-mintlify/, not this application.
Trending
Surging: 1 new project took it in the last 4 weeks, none the 4 weeks before.thin data
Who's talking about laude-institute/harbor
- A Hacker News story“Harbor – a framework for evaluating and optimizing agents and language models”Hacker News · Nov 8, 2025 · 3 points
- harbor-framework/harbor“Multiple reward keys not supported”GitHub · Feb 16, 2026 · 2 reactions
- NovaSky-AI/SkyRL“[skyrl-train][Harbor] Upstream Harbor training code to main skyrl-train”GitHub · Jan 12, 2026 · 2 reactions
- harbor-framework/harbor-datasets“Add SpreadsheetBench dataset (400 tasks)”GitHub · Feb 22, 2026 · 1 reactions
- harbor-framework/harbor“why import environment variables in environment.env in examples/configs/*.yaml”GitHub · Jan 12, 2026 · 1 reactions
The biggest 5 of 38 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 38
Talked about on GitHub and Hacker News (38 times).
On GitHub · 37
- Feb 16, 2026Multiple reward keys not supported — harbor-framework/harbor2 reactions · 1 comment
- Jan 12, 2026[skyrl-train][Harbor] Upstream Harbor training code to main skyrl-train — NovaSky-AI/SkyRL2 reactions · 1 comment
- Feb 22, 2026Add SpreadsheetBench dataset (400 tasks) — harbor-framework/harbor-datasets1 reaction
- Jan 12, 2026why import environment variables in environment.env in examples/configs/*.yaml — harbor-framework/harbor1 reaction · 5 comments
- Nov 30, 2025Add Claude Code OAuth authentication documentation — harbor-framework/harbor-docs1 reaction · 1 comment
32 more on GitHub
- Oct 5, 2026Scout pitch: Resumable, seeded benchmark runs (per-item result files, resume flag, recorded seed) — 8gi-foundation/8gent-codesays it tested it0 reactions
- Sep 26, 2026Research input for 13/14 (souls, continuity, retention): reuse existing formats, and make "forget" cascade through derived memory — BharathChinthallapalli/Test-Waters0 reactions
- Sep 15, 2026Nothing measures whether our own skills help, so every subtraction decision is a vibe check — pmgledhill102/agentic-coding-config0 reactions · 1 comment
- Aug 15, 2026Eval infrastructure: measure whether a skill rewrite still routes — Lancetnik/slop-writer0 reactions
- Aug 7, 2026docs: per-plugin CLAUDE.md with the upstream docs and local specifics — walkerhughes/agents0 reactions
- Apr 29, 2026实验记录:Qwen3-8B run-3 (msp60ius) Phase 1 OOD 跑分 — base+4 ckpt × 66 task × 3 attempts (terminus-2 + harbor + sglang) — iter215 OOD 峰 vs iter279 回落 — HansBug/OpenClaw-RL0 reactions
- Apr 10, 2026Feature: Integrate PraisonAI Agents with Terminal-Bench 2.0 (Harbor Framework) — MervinPraison/PraisonAI0 reactions · 1 comment
- Mar 1, 2026Bedrock API key env var mismatch with LiteLLM (`AWS_BEARER_TOKEN_BEDROCK`) — harbor-framework/harbor0 reactions
- Feb 21, 2026[Harbor] Proper cleanup of Trials (especially address sandbox leakage) upon RL run fails / KeyboardInterrupt — NovaSky-AI/SkyRL0 reactions
- Feb 17, 2026[Ready For Review] multi-step checkpoint trial execution — harbor-framework/harbor0 reactions · 1 comment
- Feb 16, 2026Change asyncio.gather to TaskGroup in the Harbor generator to correctly propagate cancellation signals. — NovaSky-AI/SkyRL0 reactions · 6 comments
- Feb 12, 2026Sandboxes do not get terminated after Keyboard interruption during RL — harbor-framework/harbor0 reactions · 3 comments
- Feb 12, 2026Improve startup time with installed agents — harbor-framework/harbor0 reactions · 2 comments
- Feb 11, 2026Support MCP for all agents — harbor-framework/harbor0 reactions · 2 comments
- Feb 9, 2026Stored Object: paper:url.153C303C — chtmp223/papers-feed0 reactions · 1 comment
- Feb 5, 2026Bug: breaking change from `bash -ic` to `bash -lc` — harbor-framework/harbor0 reactions · 1 comment
- Feb 5, 2026TMUX is not compatible with the alpine image in docker — harbor-framework/harbor0 reactions · 1 comment
- Jan 20, 2026OpenRouter model names not working in Harbor 0.1.42 (`--model openrouter/anthropic/claude-sonnet-4.5`) — harbor-framework/harbor0 reactions
- Jan 17, 2026Failures in agent setup are silently discarded — harbor-framework/harbor0 reactions
- Jan 16, 2026Terminus 2 Agent Uses Logger Before Initialization When Model Provider is hosted_vllm — harbor-framework/harbor0 reactions · 1 comment
- Jan 10, 2026feat: Support Agent-Trajectory Interchange Format (ATIF) for trial trajectory capture — spachava753/rollout0 reactions
- Jan 10, 2026feat(evaluation): block web access and restrict LiteLLM tools during CRS run phase — sslab-gatech/CRSBench0 reactions · 4 comments
- Jan 6, 2026docker gpu support — harbor-framework/harbor0 reactions · 11 comments
- Dec 28, 2025BUG: where does 4096 come from? — harbor-framework/harbor0 reactions · 1 comment
- Dec 24, 2025Fix: Docker environment PATH inheritance on macOS — rotemtam/harbor0 reactions
- Dec 23, 2025Adapter runs unable to find docker — harbor-framework/harbor0 reactions · 2 comments
- Dec 17, 2025Trajectory viewer for non-linear history — harbor-framework/harbor0 reactions · 2 comments
- Dec 16, 2025environment.exec(): bash: cannot set terminal process group (-1): Not a tty bash: no job control in this shell — harbor-framework/harbor0 reactions · 3 comments
- Dec 16, 2025Deniz/dc agent updates 1215 — mlfoundations/SkyRL0 reactions
- Dec 10, 2025🚨 CRITICAL: Docker Compose project name validation bug breaks Oracle on custom_docker_compose tasks — harbor-framework/harbor0 reactions · 2 comments
- Nov 30, 2025Revert litellm hack once upstream fix is released — harbor-framework/harbor0 reactions · 2 comments
- Nov 24, 2025Feature Request: Harbor as a LLM proxy — harbor-framework/harbor0 reactions · 1 comment
On Hacker News · 1
Other versions1 project uses a version of this.
Word for word · 1
- harbor-framework/harbor★ 6.2k100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 28,693 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- create-adapterSkill
- create-taskSkill
- harbor-execSkill
- publishSkill
- rewardkitSkill
- upload-parity-experimentsSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

