agentleFS
Sign inStart free
zaxbysaucezaxbysauce/opencode-swarm/.claude/skills/research-first/SKILL.md

research-first

Skill517 starsChanged 3 days agoMITView on GitHub

Apply when planning fixes, investigating tech debt, architecting solutions, or diagnosing unknown issues that may involve external frameworks, dependencies, or platform behavior. Search online for current documentation and state-of-the-art approaches before tracing through code.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 7

    Distinctive

    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:

    If the question involves external frameworks, dependencies, platform or toolchain behavior, or a suspected upstream bug: research online …

    line 22
  2. Distinctive:

    If the question is purely repo-local (project conventions, invariants, existing code structure), skip web research.

    line 24
  3. Distinctive:

    If WebSearch/WebFetch are unavailable in this environment, state that explicitly and proceed with code tracing rather than stalling.

    line 27
  4. Distinctive:

    Search current official documentation to confirm the behavior is NOT intended and NOT already fixed in a newer release

    line 32
  5. Distinctive:

    Search for state-of-the-art solutions, known community workarounds, and recent discussion of this problem type

    line 34
  6. Distinctive:

    Run multiple searches across distinct sources (official docs, issue trackers, release notes) rather than trusting the first hit

    line 36
  7. Distinctive:

    Do not stop until ≥95% confident in the root cause for every issue being investigated, and mark anything below that threshold as UNVERIFIED …

    line 41

“Common” is measured: kept by at least 6 in 10 of the 9 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 Before Planning Protocol

This skill body runs in an isolated Explore context: it has no conversation history and cannot spawn further sub-agents. Work from the task statement it receives, run searches and reads directly, and return a compact findings report.

Scope guard — research the outside world only when the problem involves it:

  • If the question involves external frameworks, dependencies, platform or toolchain behavior, or a suspected upstream bug: research online first.
  • If the question is purely repo-local (project conventions, invariants, existing code structure), skip web research; the answer lives in the repository (AGENTS.md, docs/, and the source itself).
  • If WebSearch/WebFetch are unavailable in this environment, state that explicitly and proceed with code tracing rather than stalling.

When researching:

  1. Search current official documentation to confirm the behavior is NOT intended and NOT already fixed in a newer release
  2. Search for state-of-the-art solutions, known community workarounds, and recent discussion of this problem type
  3. Run multiple searches across distinct sources (official docs, issue trackers, release notes) rather than trusting the first hit
  4. Report all findings — with URLs — before any code tracing begins

Then trace each problem end-to-end through the actual source code. Do not stop until ≥95% confident in the root cause for every issue being investigated, and mark anything below that threshold as UNVERIFIED with what evidence is missing.

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.

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.

Check your setup, free