{"id":48590025,"url":"https://github.com/lqf0624/researchstack","last_synced_at":"2026-04-08T19:01:50.726Z","repository":{"id":348471199,"uuid":"1198251399","full_name":"lqf0624/researchstack","owner":"lqf0624","description":"Workflow-first research skill pack for computer systems, networking, and AI papers.","archived":false,"fork":false,"pushed_at":"2026-04-01T09:07:15.000Z","size":70,"stargazers_count":1,"open_issues_count":0,"forks_count":1,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-01T10:30:41.677Z","etag":null,"topics":["ai4science","claude","claude-code","codex","researcher","skills"],"latest_commit_sha":null,"homepage":"","language":"JavaScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lqf0624.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-01T08:50:16.000Z","updated_at":"2026-04-01T10:04:26.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/lqf0624/researchstack","commit_stats":null,"previous_names":["lqf0624/researchstack"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/lqf0624/researchstack","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lqf0624%2Fresearchstack","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lqf0624%2Fresearchstack/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lqf0624%2Fresearchstack/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lqf0624%2Fresearchstack/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lqf0624","download_url":"https://codeload.github.com/lqf0624/researchstack/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lqf0624%2Fresearchstack/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31569400,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-08T14:31:17.711Z","status":"ssl_error","status_checked_at":"2026-04-08T14:31:17.202Z","response_time":54,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.5:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai4science","claude","claude-code","codex","researcher","skills"],"created_at":"2026-04-08T19:01:45.677Z","updated_at":"2026-04-08T19:01:50.720Z","avatar_url":"https://github.com/lqf0624.png","language":"JavaScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# researchstack\n\nEnglish | [Chinese (Simplified)](README.zh-CN.md)\n\n`researchstack` is a workflow-first research skill pack for computer systems, networking, and AI papers.\n\nIt is built for Markdown-based skill hosts such as Codex-compatible tools and Claude Code. The goal is not to give you a bag of prompts. The goal is to act like a serious paper team that can help you:\n\n- turn a vague area into a paper-sized thesis\n- stress-test novelty before you waste weeks\n- map related work against the papers that will actually be compared\n- design experiments that match the claim\n- keep runs, evidence, and artifacts disciplined\n- write, review, and rebut papers for venues such as ICLR, ASPLOS, SC, NSDI, and SIGCOMM\n\n## Why This Exists\n\nMost research assistants are good at isolated tasks. They can draft an abstract, summarize a paper, or brainstorm ideas.\n\nWhat they usually do badly is workflow:\n\n- they help write before the thesis is real\n- they suggest ideas without killing weak ones\n- they let experiment scope sprawl beyond one paper\n- they forget venue fit, reviewer risks, and prior decisions\n\n`researchstack` is built to push in the opposite direction. It routes work through a stricter path:\n\n`idea -\u003e review -\u003e literature -\u003e experiment -\u003e artifact -\u003e writing -\u003e submission -\u003e rebuttal`\n\n## What You Get\n\n- A root router skill: `researchstack`\n- 17 focused research skills with `researchstack-*` names for easier search\n- Source skills for Claude Code style installs\n- Generated `.agents/skills/` distribution for Codex-compatible installs\n- A local memory layer for long-running projects\n- Templates for experiment ops, reproduction, paper layout, figures, and submission gating\n- Example workflows and a realistic example project\n\nHost UIs group the skills under `Researchstack: ...` display names so they are easier to find in long skill lists.\n\n## Skill Catalog\n\n### Routing and Memory\n\n- `researchstack`\n  Root router. Use when the request spans multiple stages and you want the pack to choose the right research workflow.\n- `researchstack-learn`\n  Project memory and researcher preference management. Use to store durable decisions such as thesis boundaries, venue choices, reviewer risks, and evaluation rules.\n\n### Idea Formation\n\n- `researchstack-idea-finder`\n  Start here when you only have a broad area or interest, not a paper-sized idea. It scans recent papers, proposes candidate topics, runs them through mandatory review, and only returns surviving one-paper thesis candidates.\n- `researchstack-lab-intake`\n  Turn a rough direction into a paper brief. Good for venue targeting, thesis framing, assumptions, threat model, and immediate next steps.\n- `researchstack-idea-review`\n  Skeptical research-idea review. Use to judge novelty, importance, scope, evidence burden, and likely reviewer objections.\n- `researchstack-idea-refine`\n  Guided repair loop for ideas that are interesting but not yet publishable.\n- `researchstack-literature-map`\n  Related-work and positioning skill. Use to find the closest comparison set, novelty boundary, and competing narratives.\n\n### Existing Paper Reproduction\n\n- `researchstack-paper-reproduction`\n  Start from a paper PDF, appendix, or repo. Reconstruct the method, extract claims and experiments, identify ambiguity, and produce a reproduction verdict.\n\n### Evidence and Implementation\n\n- `researchstack-experiment-design`\n  Build the evaluation matrix: baselines, metrics, workloads, ablations, scaling checks, and failure tests.\n- `researchstack-experiment-ops`\n  Keep runs disciplined: runbooks, logging, seed control, checkpointing, result triage, and evidence promotion.\n- `researchstack-artifact-audit`\n  Trace every figure and claim back to scripts, configs, seeds, hardware assumptions, and outputs.\n- `researchstack-code-review`\n  Review research code for correctness, hidden bias, invalid conclusions, and claim-implementation mismatch.\n\n### Writing and Presentation\n\n- `researchstack-paper-write`\n  Draft or revise sections with claim-evidence discipline.\n- `researchstack-figure-studio`\n  Plan figures and tables that carry evidentiary weight instead of decorative value.\n- `researchstack-paper-layout`\n  Improve LaTeX structure, captions, notation, tables, appendix flow, and camera-ready hygiene.\n\n### Submission and Review\n\n- `researchstack-submission-gate`\n  Final go/no-go check. Use to decide whether to submit now, delay, or retarget.\n- `researchstack-peer-review`\n  Simulate strict conference reviewers and likely rejection paths.\n- `researchstack-rebuttal-coach`\n  Turn reviews into a disciplined rebuttal plan and point-by-point response strategy.\n\n## Typical Workflows\n\n### 1. You only have a broad area\n\nUse:\n\n1. `researchstack-idea-finder`\n2. `researchstack-lab-intake`\n3. `researchstack-idea-review`\n4. `researchstack-literature-map`\n\nOutcome:\n\n- a few reviewed topic cards\n- rejected directions with reasons\n- one-paper thesis candidates\n- a clean handoff into real paper planning\n\n### 2. You have a rough paper idea already\n\nUse:\n\n1. `researchstack-lab-intake`\n2. `researchstack-idea-review`\n3. `researchstack-idea-refine` if needed\n4. `researchstack-literature-map`\n\nOutcome:\n\n- venue target\n- one-sentence thesis\n- novelty boundary\n- highest-risk unknowns\n\n### 3. You already have code and results\n\nUse:\n\n1. `researchstack-experiment-design`\n2. `researchstack-experiment-ops`\n3. `researchstack-artifact-audit`\n4. `researchstack-code-review`\n\nOutcome:\n\n- claim-to-evidence matrix\n- controlled run process\n- reproducibility gaps\n- implementation risks that could sink the paper\n\n### 4. You want to reproduce an existing paper\n\nUse:\n\n1. `researchstack-paper-reproduction`\n2. `researchstack-experiment-design`\n3. `researchstack-experiment-ops`\n4. `researchstack-artifact-audit`\n5. `researchstack-peer-review`\n\nOutcome:\n\n- reproduction brief\n- ambiguity log\n- reconstruction matrix\n- reviewer-style critique of reproducibility\n\n### 5. You are writing or polishing a submission\n\nUse:\n\n1. `researchstack-paper-write`\n2. `researchstack-figure-studio`\n3. `researchstack-paper-layout`\n4. `researchstack-submission-gate`\n5. `researchstack-peer-review`\n6. `researchstack-rebuttal-coach`\n\nOutcome:\n\n- tighter sections\n- stronger figures and captions\n- cleaner layout\n- a clearer submit-or-delay decision\n\n## Memory Layer\n\n`researchstack` includes a lightweight local memory layer so long-running projects do not restart from zero every session.\n\n- Project memory: `~/.researchstack/projects/\u003cslug\u003e/memory.jsonl`\n- Researcher preferences: `~/.researchstack/profile/preferences.json`\n\nUse it for:\n\n- thesis boundaries\n- venue decisions\n- baseline and evaluation rules\n- hardware or dataset constraints\n- recurring reviewer risks\n- repeated writing weaknesses\n\nUse this helper to initialize a stable slug and memory layout:\n\n```bash\nbun run memory:init -- --root \u003crepo-root\u003e\n```\n\nSlug rule:\n\n- `\u003crepo-name\u003e-\u003csha256(normalized-remote-or-root-path)[:6]\u003e`\n\n## Install\n\nRequirements:\n\n- Git\n- Bun\n- Git Bash is recommended on Windows for running `./setup`\n- WSL only works if your host also scans the WSL-side skills directory, or if you use `--target` to point at a Windows-visible skills path\n\nRepository:\n\n- GitHub: [lqf0624/researchstack](https://github.com/lqf0624/researchstack)\n\n### Claude Code\n\nInstall globally:\n\n```bash\ngit clone https://github.com/lqf0624/researchstack.git ~/.claude/skills/researchstack\ncd ~/.claude/skills/researchstack\n./setup --host claude\n```\n\nInstall into one repo:\n\n```bash\ncp -Rf ~/.claude/skills/researchstack .claude/skills/researchstack\nrm -rf .claude/skills/researchstack/.git\ncd .claude/skills/researchstack\n./setup --host claude\n```\n\n### Codex-Compatible Hosts\n\nInstall globally:\n\n```bash\ngit clone https://github.com/lqf0624/researchstack.git ~/researchstack\ncd ~/researchstack\n./setup --host codex\n```\n\nInstall into one repo:\n\n```bash\ngit clone https://github.com/lqf0624/researchstack.git .agents/skills/researchstack\nrm -rf .agents/skills/researchstack/.git\ncd .agents/skills/researchstack\n./setup --host codex\n```\n\nWhen setup runs from `.agents/skills/researchstack`, it keeps the vendored checkout as the root `researchstack` skill, installs generated `researchstack-*` sibling skills next to it, and does not write to `~/.codex/skills`.\n\nExplicit target path:\n\n```bash\ncd ~/researchstack\n./setup --host codex --target /path/to/project/.agents/skills\n```\n\nAuto-detect host from a shared checkout:\n\n```bash\n./setup --host auto\n```\n\n## Repository Layout\n\n- [SKILL.md](SKILL.md)\n  Root router and workflow entrypoint.\n- `artifact-audit/`, `idea-review/`, `paper-write/`, and similar folders\n  Source skills.\n- [references](references)\n  Shared policy and review references.\n- per-skill `assets/`\n  Templates and reusable checklists.\n- per-skill `agents/openai.yaml`\n  Host-facing UI metadata.\n- [scripts/gen-skill-docs.js](scripts/gen-skill-docs.js)\n  Generated distribution builder.\n- [scripts/init-memory.js](scripts/init-memory.js)\n  Memory slug/bootstrap helper.\n- [setup](setup)\n  Installer for Claude and Codex-compatible hosts.\n- [.agents/skills](.agents/skills)\n  Generated distribution output.\n\n## Demos and Examples\n\n- [docs/demo-workflow-gpu-simulator-paper.md](docs/demo-workflow-gpu-simulator-paper.md)\n  End-to-end example around a GPU simulator paper.\n- [docs/demo-workflow-paper-reproduction.md](docs/demo-workflow-paper-reproduction.md)\n  Reproduction branch demo from a published paper.\n- [docs/demo-memory-flow.md](docs/demo-memory-flow.md)\n  Memory flow demo across repeated sessions.\n- [examples/nsdi-tail-latency-paper/README.md](examples/nsdi-tail-latency-paper/README.md)\n  Realistic example project artifacts.\n\n## Contributing\n\nSee:\n\n- [CONTRIBUTING.md](CONTRIBUTING.md)\n- [LICENSE](LICENSE)\n\n## Troubleshooting\n\n- Skills not showing up after install: rerun `./setup` and restart the host app so it rescans the skills directory.\n- Repo-local install ended up in your home directory: use the documented gstack-style paths, or rerun with `--target` if your host uses a custom skills directory.\n- Windows + WSL install succeeded but the desktop app shows nothing: rerun from Git Bash, or use `./setup --target \u003cWindows-visible-skills-dir\u003e` so the files land where the app actually scans.\n- You want generated distribution output without a global install: run `bun run gen:skill-docs --host codex` and use `.agents/skills/` directly.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flqf0624%2Fresearchstack","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flqf0624%2Fresearchstack","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flqf0624%2Fresearchstack/lists"}