Agent Skills

jackin-research

Use only when the user explicitly requests this skill. Produces a standalone multi-page research dossier on the jackin❯ docs site, drawing on web and codebase evidence.

Install

npx skills add https://github.com/jackin-project/jackin-dev --skill jackin-research
SKILL.md

jackin-research

Produce a standalone research dossier — a multi-page deliverable published on the jackin❯ docs site — for an open question or roadmap topic. Brief-driven: author the brief, then execute it. Gathers and synthesizes evidence (web + codebase); does not make product design decisions.

Reference implementation: docs/content/docs/research/token-optimization-research/ in the jackin❯ repo.

When to use

  • Operator runs jackin-research <slug> for a substantial investigation.

When NOT to use

  • Design decisions on a roadmap item → jackin-brainstorm.
  • Quick lookup → deep-research or Explore directly, no dossier.

Arguments

  • --brief-only — author prompt.mdx (the brief) and stop; execute later via /goal Follow <brief>.
  • --in-roadmap — store findings inline in the roadmap item instead of a separate dossier (small research only).
  • --web-only / --codebase-only — restrict gathering to one pass.

Output layout (default: separate folder)

docs/content/docs/research/<slug>/
├── meta.json     # { title, defaultOpen:false, pages:[...] } — sidebar order
├── index.mdx     # dossier landing: headline numbers, how-to-read, tier list
├── prompt.mdx    # the brief that was run (the spec; carries the /goal run line)
├── NN-*.mdx      # numbered chapters (00 summary, 01.. foundations, 10-20 areas, 30+ synthesis)
└── tools/        # optional scripts + own meta.json + index.mdx

Also add <slug> to the parent docs/content/docs/research/meta.json pages. Big research → own folder (default); small → --in-roadmap.

Process

  1. Scope / brief. From the topic (or roadmap item), draft or load prompt.mdx: mission, chapter list, evidence rules. With --brief-only, stop here.
  2. Gather. Web via the built-in deep-research; codebase via Explore/grep. Every external claim carries a source URL; every local number carries its method.
  3. Write. index.mdx (headline numbers + how-to-read + tier list) and the numbered chapters in a fixed per-technique record schema; bundle any reproduction scripts under tools/.
  4. Wire the sidebar. Create/update the dossier meta.json and the parent research/meta.json.
  5. Docs gate + commit. Run the bun docs gate (build, check:repo-links, tsc --noEmit, test) from docs/. Commit docs(research):; push.

Common mistakes

  • Forcing a Problem/Why/Design shape — a dossier is free-form, not a roadmap item.
  • Making design decisions — that is brainstorm.
  • An unsourced external claim, or a local number without its method.
  • meta.json pages out of sync with the files on disk.

Tooling

cargo xtask research scaffold <slug> creates the folder + meta.json; cargo xtask research check validates pages against disk. Execution typically runs via the external /goal Follow <brief>.

Related skills

researchmattpocock575KInvestigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.paper-context-resolverlllllllama451KRigor Paper Context helper for README-first deep learning repo reproduction. Use only when the README and repository files leave a narrow reproduction-critical gap and the task is to resolve a specific paper detail such as dataset split, preprocessing, evaluation protocol, checkpoint mapping, or runtime assumption from primary paper sources while recording conflicts. Do not use for general paper summary, repo scanning, environment setup, command execution, title-only paper lookup, or replacing Renv-and-assets-bootstraplllllllama450KRigor Setup skill for README-first deep learning repo reproduction. Use when the task is specifically to prepare a conservative conda-first environment, checkpoint and dataset path assumptions, cache location hints, and setup notes before any run on a README-documented repository. Do not use for repo scanning, full orchestration, paper interpretation, final run reporting, or generic environment setup that is not tied to a specific reproduction target.ai-research-explorelllllllama311KRigor Explore compatible skill slug for meaningful and potentially novel deep learning research candidates. Use when the researcher has chosen the task family, dataset, benchmark, evaluation method, provided SOTA references, and wants candidate-only exploration on top of `current_research` with auditable repo understanding, idea gating, fair comparison, and governed experiments written to `explore_outputs/`. Do not use for README-first trusted reproduction, open-ended direction finding, narrow c

Search skills and MCP servers

Fuzzy search across 23,137 skills and servers