grounded-answer
Skill0 starsChanged 7 days agoMITView on GitHub
Prevents fabrication (hallucination) when answering from documents or facts. Use when answering questions about provided files, PDFs, transcripts, contracts or data; when summarizing sources; when the user asks for citations, statistics, quotes, references or "facts"; or whenever accuracy matters more than fluency. Quotes evidence first, cites every claim, marks unknowns, and removes unsupported claims.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- JetXu-LLM/DocMasontheir own version, first seen Oct 8, 2026
- itsmeyaw/skillstheir own version, first seen Sep 2, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- JetXu-LLM/DocMason★ 152same name
- itsmeyaw/skills★ 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 repository11 other files it gives its agents.
- drop-in/AGENTS.mdAGENTS.md
- llms.txtllms.txt
- clarify-firstSkill
- coverage-checkSkill
- fresh-eyes-reviewSkill
- fresh-start-handoffSkill
- honest-critiqueSkill
- learning-tutorSkill
- prompt-improverSkill
- task-specSkill
- verify-claimsSkill
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
- Distinctive: line 16
Say what's missing: If the material doesn't answer part of the question, write "Not in the provided material" for that part.
- Distinctive: line 22
Only cite sources you actually opened in this session.
- Distinctive: line 23
If you can't verify that a source exists and says what you claim, say so instead of citing it.
≈ “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.
Files
Grounded answer
Models hallucinate partly because training and evaluation reward confident guessing over "I don't know". Counter that explicitly: ground every claim in evidence and make abstaining acceptable.
Steps
- Collect evidence first. From the provided material (or from sources you actually retrieved with a search tool), extract the passages that bear on the question. Quote them word for word, numbered, with their location:
[1] "…" (report.pdf, p. 12)
- Answer only from the evidence. Attach a citation like [1] to every factual claim. If you add general knowledge, label it [general knowledge].
- Say what's missing. If the material doesn't answer part of the question, write "Not in the provided material" for that part. Don't fill gaps with plausible guesses.
- Retract what you can't support. Re-read your draft; for each claim, check that a quoted passage supports it. Delete any claim without support and note [removed: unsupported].
- Flag confidence. Where it helps, tag claims as [from source], [from memory, unverified] or [inference].
Citations and references
- Only cite sources you actually opened in this session. Never construct a URL, DOI, case name, page number or quote from memory.
- If you can't verify that a source exists and says what you claim, say so instead of citing it.
Numbers
- Compute totals, percentages and statistics by running code, and show the code.
Evidence
- Why models guess instead of abstaining: Kalai et al. (OpenAI), 2025 — https://arxiv.org/abs/2509.04664
- Allow "I don't know", extract quotes first, cite or retract: Anthropic, Reduce hallucinations — https://platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-hallucinations
- 2026-released models still fabricated references in 35% of responses (preprint): Topaz et al., 2026 — https://arxiv.org/abs/2609.14988
- AI-fabricated citations in court filings: 2,266 cases as of 2026-11-05 — https://www.damiencharlotin.com/hallucinations/
- More in the guide: https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/04-failure-modes.md#3-fabrication-hallucination
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.

