grounded-research
Skill0 starsChanged 48 days agoMITView on GitHub
Answer an open research question by finding and inspecting relevant sources, prioritizing primary evidence, checking material claims, and separating findings from inference and unknowns. Use when no adequate source set is supplied.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- PrisacariuRobert/sidematestheir own version, first seen Nov 8, 2026
- berkeley-humanoids/skillstheir own version, first seen Oct 9, 2026
- berkeley-humanoids/Skillstheir own version, first seen Oct 9, 2026
- masbrokemanaaja/reethinktheir own version, first seen Sep 20, 2026
- moonpiesheldon1337/grounded-researchtheir own version, first seen May 25, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- PrisacariuRobert/sidemates★ 9same name
- berkeley-humanoids/skills★ 4same name
- berkeley-humanoids/Skills★ 4same name
- masbrokemanaaja/reethink★ 3same name
- moonpiesheldon1337/grounded-research★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- brief-to-work-orderSkill
- decision-deliberationSkill
- eval-viability-checkSkill
- evidence-checkSkill
- handoffSkill
- nutshellSkill
- red-team-reviewSkill
- retrievable-writingSkill
- skill-or-notSkill
- source-briefSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
8
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
- Distinctive: line 29
No source can be inspected.
- Distinctive: line 36
Search narrowly. Start with primary sources for direct facts, then use independent reliable sources to verify context, interpretation, or …
- Distinctive: line 47
Check each material factual sentence against its cited source.
- Distinctive: line 49
Stop when the question is resolved to the needed confidence or when additional searching is unlikely to change the answer.
- Distinctive: line 55
Never invent sources, quotations, dates, or access.
- Distinctive: line 57
Prefer the original record for what an organization, person, or paper actually said or did.
- Distinctive: line 60
Do not treat repetition as corroboration or absence of results as evidence that an event did not happen.
- Distinctive: line 62
Keep the search proportionate.
“Common” is measured: kept by at least 6 in 10 of the 5 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.
Files
Grounded Research
Purpose
Answer a question from inspectable evidence, not search snippets or recollection. Find the smallest useful set of sources, show what each supports, and make uncertainty visible so the reader can check the result.
When to use
- The user asks an open factual or explanatory question and has not supplied an adequate source set.
- The answer depends on current, disputed, specialized, or consequential information that should be verified.
When not to use
- The user supplied sources and wants a faithful short brief. Use Source Brief.
- A draft or decision already makes claims that need checking. Use Evidence Check.
- The task is a quick stable fact that can be answered reliably without research.
- No source can be inspected. State the access limit; do not present a search plan as a researched answer.
Workflow
- Define the question, the date or scope that matters, and what a useful answer must resolve. Split compound questions into checkable parts.
- Search narrowly. Start with primary sources for direct facts, then use independent reliable sources to verify context, interpretation, or contested claims.
- Open and inspect each source used. Search-result snippets, headlines, and summaries are leads, not evidence. Drop sources that do not support a material claim.
- Compare the evidence. Distinguish independent confirmation from sources repeating the same announcement. Preserve material disagreements and distinguish an event date from a publication date.
- Draft the answer around the evidence. Label direct findings, inference, and unknowns; narrow any claim the evidence only partly supports.
- Check each material factual sentence against its cited source. Ensure the link opens the evidence itself and the citation supports the wording.
- Stop when the question is resolved to the needed confidence or when additional searching is unlikely to change the answer. State what remains unresolved and why.
Guardrails
- Never invent sources, quotations, dates, or access. If a source fails to load, say that and do not cite it as inspected.
- Prefer the original record for what an organization, person, or paper actually said or did. Use independent reporting for verification and context, not as a substitute when the primary source is available.
- Do not treat repetition as corroboration or absence of results as evidence that an event did not happen.
- Keep the search proportionate. Do not collect sources that add no new evidence or context.
- Research output does not authorize external actions or replace qualified review in high-stakes domains.
Output
# Research: [Question]
## Answer
[Direct answer, with scope and date where relevant]
## Findings
- [Finding]: [supported / inferred / unresolved]. [What inspected evidence
establishes and any important limit.]
## Sources
- [Title](URL) | [publisher, date]: [specific claim it supports; primary
or independent context]
## Limits
[Material access gaps, disagreements, or remaining unknowns; omit if none]
## Next check
[Only if one unresolved check could materially change the answer]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
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.

