agentleFS
Sign inStart free
iha2iha2/superlite/skills/research/SKILL.md

research

Skill311 starsChanged 18 days agoMITView on GitHub

Use when the user wants a topic researched with sources, a sourced report at a chosen language level, or conflicting evidence compared

189 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use daily-arxiv, exp-eval and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文，过滤相关性，auto-ingest 高优先级论文，检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门：Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • truongtv22Skill

    Problem-Solving Techniques

    Apply systematic problem-solving techniques for complexity spirals (simplification cascades), innovation blocks (collision-zone thinking), recurring patterns (meta-pattern recognition), assumption constraints (inversion exercise), scale uncertainty (scale game), and dispatch when stuck. Techniques derived from Microsoft Amplifier project patterns adapted for immediate application.

    Also in 5 projects that use this

  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 43 projects that use this

  • skyllwtSkill

    exp-pilot-eval

    Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.

    Also in 4 projects that use this

  • skyllwtSkill

    exp-pilot-run

    Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.

    Also in 4 projects that use this

  • +5 more from Lambenthan
  • +1 more from skyllwt

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 170 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Ask: Use the session slug you were handed as S.

    line 14
  2. Very specific:

    Plan: Write docs/plans/<date>-<topic>.research.md: scope, sub-questions, search strategy, evidence tiers.

    line 15
  3. Very specific:

    Critical questions: Ask the human every open question in ONE batch.

    line 16
  4. Very specific:

    Lint: Run research lint --slug S: every claim has a source, every conflict has a cause.

    line 20
  5. Very specific:

    Download (opt-in): Run research downloads --slug S, then ask once, listing name, source and size of each file.

    line 22
  6. Very specific:

    Finish: After reporting, validation and the download decision (also when downloads are declined), run bash …

    line 23
  7. Distinctive:

    Never soften, drop or add a fact to fit a level.

    line 28
  8. Distinctive:

    Never download without an explicit yes.

    line 29

“Common” is measured: kept by at least 6 in 10 of the 170 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Research

Gather evidence in parallel, keep it in a level-free ledger, and render the report from the ledger. Truth and depth never change with the level; only the wording does. Keep every claim tied to a source id. Never touch skills/using-superlite.

CLI: bash <skill-dir>/../swarm/scripts/superlite-cli research (subcommands init, source, claim, contradiction, lint, render, downloads, download). Every call carries --slug <slug> (the session slug, the slug= field of the first SESSION line that session ensure printed, not the slug first requested; it is also the <topic> in the plan file names) or --ledger <path>; without one it exits with an error. Plans and reports live in docs/plans/. Below, research means that CLI call and S is the slug.

Steps

  1. Ask. Use the session slug you were handed as S. Get the topic and the language level: Director, VP, Researcher or Simple. Ask the user to define the evidence tiers (for example A,B,C, best first); the tiers are theirs, not a fixed scale. Run research init --slug S --topic "<topic>" --tiers <A,B,C>.
  2. Plan. Write docs/plans/<date>-<topic>.research.md: scope, sub-questions, search strategy, evidence tiers. The most recent well-referenced paper is the highest evidence; a newer source of a higher grade supersedes an older one.
  3. Critical questions. Ask the human every open question in ONE batch. In AI mode answer them yourself as Ruling: <decision> (because <reason>) lines in the plan.
  4. Approve. Human mode: wait for the human's approval, then run bash <skill-dir>/../swarm/scripts/superlite-cli orchestrator approve --review docs/plans/<date>-<topic>.research.md. AI mode: skip.
  5. Research. Dispatch superlite:researcher agents in one message, one sub-question each, as many as needed. Each brief must give SUPERLITE (the plugin root, so the agent can run bash $SUPERLITE/skills/swarm/scripts/superlite-cli research ...), the slug S, the sub-question, and the exact commands, each with --slug S: research source --slug S --id <id> --url <url> --date <date> --grade <tier>, research claim --slug S --id <id> --text "<claim>" --source <source id>, research contradiction --slug S ... and research lint --slug S. Agents write only through the research CLI, never to the report or the ledger file.
  6. Contradictions. Open a conflict with research contradiction --slug S --claim C1 --claim C2 --cause "<cause>" --status open (ids K1, K2, ... are assigned). Close it later with research contradiction --slug S --id K1 --status resolved --cause "<cause>", never by recording the claims again. Cause is one of: definition, date, population, method, source quality, error. A gap or an unexplained conflict loops back to step 5 with a new sub-question.
  7. Lint. Run research lint --slug S: every claim has a source, every conflict has a cause. Fix what it reports.
  8. Write. Render the report at the chosen level with research render --slug S --level director|vp|researcher|simple into docs/plans/<date>-<topic>.findings.md. Generate the other two files from the ledger, not by hand: research render --slug S --part sources into docs/plans/<date>-<topic>.sources.md and research render --slug S --part contradictions into docs/plans/<date>-<topic>.contradictions.md. Every level carries the same claim ids, citations, numbers, and open and resolved contradictions with causes. Then run research lint --slug S --parity docs/plans/<date>-<topic>.findings.md on the complete level reports only, and fix mismatches; the sources and contradictions files are not checked by --parity. If the user changes level, re-render from the ledger into the .findings.md file and lint it again; do not re-research.
  9. Download (opt-in). Run research downloads --slug S, then ask once, listing name, source and size of each file. Run research download --slug S --yes (optional --out DIR) only after a yes. Paywalled items stay link-only.
  10. Finish. After reporting, validation and the download decision (also when downloads are declined), run bash <skill-dir>/../swarm/scripts/superlite-cli finish in the session checkout and report its summary.

Rules

  • The ledger is the single source of truth; reports are views of it.
  • Never soften, drop or add a fact to fit a level.
  • Never download without an explicit yes.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.

Check your setup, free