Skip to content
Skills Library
Science & research · Analysis

doubt

/doubt

Turn a contested question, research brief, technical choice, or high-impact decision into a source-grounded interactive evidence map. Use when an agent must preserve support, contradiction, qualification, and missing evidence; make every evidence edge traceable to a dated source region; pressure-test a conclusion; or render a shareable HTML decision artifact instead of collapsing disagreement into a prose summary.

Repository
doubt ↗
Licence
MIT
Last repository change
31.07.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
6
Scripts
2
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 31.07.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What to check before installing
  • 2 scripts in this package. Review them before installing, as Claude may execute them.

Scripts: scripts/contract.mjs, scripts/validate.mjs

What it does and when to use it

Turn a contested question, research brief, technical choice, or high-impact decision into a source-grounded interactive evidence map. Use when an agent must preserve support, contradiction, qualification, and missing evidence; make every evidence edge traceable to a dated source region; pressure-test a conclusion; or render a shareable HTML decision artifact instead of collapsing disagreement into a prose summary.

Original author description

Turn a contested question, research brief, technical choice, or high-impact decision into a source-grounded interactive evidence map. Use when an agent must preserve support, contradiction, qualification, and missing evidence; make every evidence edge traceable to a dated source region; pressure-test a conclusion; or render a shareable HTML decision artifact instead of collapsing disagreement into a prose summary.

How it works

  1. Workflow
  2. Quality gates
  3. Do not
Original SKILL.md headings · 95 lines, 689 words
SKILL.md preview
---
name: doubt
description: Turn a contested question, research brief, technical choice, or high-impact decision into a source-grounded interactive evidence map. Use when an agent must preserve support, contradiction, qualification, and missing evidence; make every evidence edge traceable to a dated source region; pressure-test a conclusion; or render a shareable HTML decision artifact instead of collapsing disagreement into a prose summary.
---

# Doubt

Produce a decision artifact that shows why a conclusion holds, what pushes
against it, and what remains unknown. Do not use a graph to decorate an answer
that has not been sourced.

## Workflow

1. **Frame one decision.** Write a falsifiable question and one provisional
   position. Narrow broad topics until a reader can tell what action or belief
   the map is testing.
2. **Collect current source regions.** Prefer direct observations and primary
   sources. For each item, preserve the URL or local path, publisher, date,
   retrieval date, section/page/line locator, and a short verbatim excerpt copied
   from that source region. Never put an agent-written paraphrase in `excerpt`;
   keep interpretation in the evidence node's `text`. Read
   [references/evidence-ladder.md](references/evidence-ladder.md)
   when source quality is disputed.
3. **Atomize the reasoning.** Create only four node types:
   - `position`: the single current verdict;
   - `claim`: an intermediate proposition;
   - `evidence`: a sourced observation;
   - `unknown`: a specific missing fact that could change the result.
4. **Type every edge.** Use `supports`, `contradicts`, `qualifies`, or
   `missing`. Add a plain-language note explaining why the source node bears on
   the target. Similarity is not support; opposite wording is not necessarily
   contradiction.
5. **Preserve disagreement.** Do not delete contrary evidence merely because
   the verdict survives it. Represent scope differences as `qualifies`, not
   false contradictions.
6. **Refuse false precision.** Do not invent confidence percentages. Express
   uncertainty as an explicit unknown, a qualifying edge, or a narrower
   verdict.
7. **Write the map JSON.** Follow
   [references/map-schema.md](references/map-schema.md). Keep node IDs short,
   stable, and semantic.
8. **Fail closed offline, explicitly verify, then render.** The host exposes
   the path of this `SKILL.md` when it loads the skill. Resolve
   `scripts/validate.mjs` relative to that directory and replace
   `<skill-dir>` below with the concrete path; do not pass the placeholder
   literally.

   ```bash
   node <skill-dir>/scripts/validate.mjs decision.doubt.json
   # Required before calling an externally sourced map finished or verified;
   # run only after the user explicitly permits source retrieval:
   npx doubt-ai verify decision.doubt.json --out decision.verified.doubt.json
   npx doubt-ai map decision.verified.doubt.json --out decision.html
   ```

   The bundled validator uses only Node.js 18+ built-ins and must work without
   package installation or network access. Fix every validation finding. Never
   bypass missing-source, unused-evidence, unsupported-position, or
   false-precision findings. Only report validation success and a receipt after
   the validator exits zero and prints `VALID` followed by a 64-character
   receipt (or JSON with `valid: true` and that receipt). A file hash, node
   count, or syntactic JSON check is not a Doubt receipt. If the validator
   cannot run, report the run as blocked; do not claim success. Source
   verification is a required final gate for a finished externally sourced map,
   but do not run `verify` implicitly: it can make outbound requests for
   recorded HTTP sources. Ask for permission first. If permission is denied or
   unavailable, you may render the original JSON only as a recorded-only draft;
   do not call it finished or verified. Local file verification does not use the
   network. If verification fails, preserve the mismatch or retrieval failure
   instead of writing an attestation.
9. **Inspect the HTML.** Verify that the question, verdict, evidence cards,
   contradictions, unknowns, edge focus, filters, and source links are readable.
   The JSON remains the canonical editable artifact.

## Quality gates

A finished map must satisfy all of these:

- exactly one position has incoming reasoning;
- every evidence node names a source and participates in an edge;
- every source has a publication date, retrieval date, bounded region, and is used;
- every source excerpt is verbatim text from its bounded region, never a summary;
- every non-position node has a directed path to the position, without cycles;
- the map contains contrary or qualifying evidence when the source set contains
  it;
- unresolved decision-changing facts appear as `unknown` nodes;
- every edge note states an entailment, contradiction, qualification, or gap;
- the verdict is no broader than the evidence.

## Do not

- Treat an agent-written summary as evidence for itself.
- Attach a citation that merely mentions the topic.
- Generate decorative orphan nodes.
- Infer that two sources conflict when populations, versions, dates, or
  evaluation conditions differ.
- Hide a failed retrieval behind authoritative prose.
- Replace a simple, settled fact with an unnecessary map.
Files in this package (6)
  • SKILL.md 5 KB
  • agents/openai.yaml 1 KB
  • references/evidence-ladder.md 2 KB
  • references/map-schema.md 3 KB
  • scripts/contract.mjs 18 KB
  • scripts/validate.mjs 4 KB

Versions

RepositoryScoreLicence
This page · alsoleg89/doubt100MIT
htlin222/dotfiles100MIT

Identical content appears in 2 other GitHub folders. We show one copy with the highest score.

Database snapshot: 13.11.2026 · package version 72feb23102e4 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100build-evidence-map
    Build an auditable evidence map for a contested technical choice, research synthesis, proposal review, or consequential decision. Use when Copilot must preserve supporting, contradicting, qualifying, and missing evidence with exact source regions instead of collapsing disagreement into prose.
  • 90research-evidence-map
    Turn a technical research question into a traceable evidence map with source quality, uncertainty, and actionable next experiments. Use for model architecture choices, benchmark claims, library selection, and unfamiliar technical topics.
  • 90research-brief
    Research a question and write a short, sourced brief. Use when the user says "research", "look into", "find out", "what does the evidence say", "is it true that", or needs a fact base before a decision.
  • 100research
    Produce a source-grounded, cited HTML research report on a question. Gathers a brief (depth, recency, source mix, audience) then launches a bundled Dynamic Workflow that fans out firecrawl retrieval into findings, gates rounds on coverage + contradictions, synthesizes a cited…
  • 100source-backed-research-brief
    Research a defined question and produce a concise, decision-ready brief with traceable current sources, explicit uncertainty, conflicting-evidence handling, and separated facts and inference. Use for market, product, policy, vendor, technical, or other web-backed research; not…
  • 90research-evidence
    Build a traceable map of claims, original sources, contradictions, and evidence gaps for a research question. Use when gathering literature or checking what retrieved material actually supports.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.