agentleFS
wrencanflywrencanfly/research-loop-skills/skills/research-frame/SKILL.md

research-frame

Skill49 starsChanged 8 days agoMITView on GitHub

Turn a broad research objective into an answerable question, scope, evidence standard, and bounded plan. Use at the start of an investigation or when its question changes.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the user's objective, available materials, intended decision or audience, and existing constraints.

    line 12
  2. Distinctive:

    Computational investigation: test a proposition using accessible data, code, simulation, or formal reasoning.

    line 20
  3. Distinctive:

    Use both when needed; a literature result and an experimental result remain distinct evidence.

    line 21

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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.

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

Research Frame

Define what would make this investigation useful and what would count as an answer. Work in the user's language. Preserve an existing good brief instead of restarting it.

Inputs

Use the user's objective, available materials, intended decision or audience, and existing constraints. Infer routine choices from context; ask only for missing information that materially changes the research question or permitted work. Continue independent preparation while waiting.

Procedure

  1. Express the main question so a reader could distinguish an answer from background information. State the population or system, comparison, outcome, and time window when relevant.
  2. Explain why answering it matters. Separate research value from ease of automation. If the request is broad, choose a bounded first question that can change the user's decision.
  3. Select the work mode:
    • Evidence synthesis: resolve what existing material supports, where it disagrees, and what remains unknown. A systematic review requires an explicit search and screening method; do not imply exhaustiveness from a few searches.
    • Computational investigation: test a proposition using accessible data, code, simulation, or formal reasoning. State which claims need real-world validation beyond computation.
    • Use both when needed; a literature result and an experimental result remain distinct evidence.
  4. List consequential assumptions and plausible alternative answers. Name the evidence that would change the initial view. For descriptive questions, use competing interpretations rather than forcing a statistical hypothesis.
  5. Identify the current bottleneck: an unclear question, missing source, unavailable data, unreliable method, slow execution, or uncertain interpretation. Choose the next action for the information it can add.
  6. Bound the work using the user's budget. If none exists, state a modest initial search or experiment batch and a stopping rule. Do not interpret silence as permission for ongoing expense or external action.
  7. Assign autonomy per action based on available inputs, an observable correctness check, error consequences, and existing authorization. Continue authorized, reversible work. Turn inaccessible physical or interpersonal work into a concrete handoff.

Output

Produce a compact brief, in prose or the project's existing format:

  • Main question and intended use of the answer.
  • Scope, exclusions, work mode, and known inputs.
  • Initial alternatives and assumptions.
  • Evidence standard and what would change the conclusion.
  • Next action, verification method, budget, and stopping conditions.
  • Dependencies or decisions requiring human input, if any.

Useful stopping conditions include an answer meeting the evidence standard, budget exhaustion, repeated searches yielding no material new evidence, or a necessary input being unavailable. An unresolved conclusion is a valid outcome.

Completion check

A different researcher should be able to tell which work is in scope, why the next action matters, and how its result will be judged. If the question cannot be answered with available evidence, state the narrower answerable question and preserve the original as unresolved.

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.