How do I install the Evidence Gathering skill?

Run npx skillmds@latest add npbuilds/evidence-gathering in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Evidence Gathering safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Evidence Gathering?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Evidence Gathering free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Evidence Gathering?

npbuilds (@npbuilds) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Evidence Gathering by npbuilds — Agent Skill

Name: Evidence Gathering
Author: npbuilds

Evidence Gathering

Route a research question to the evidence-acquisition skills — claim decomposition, source triangulation, funding/COI tracing, paywall strategy, and the generative agentic-researcher. Activate during spelunker Phases 2–3 (Decompose → Investigate) when a question needs to be broken into claims and each claim backed by independent sources. This director owns the "find and qualify the evidence" half of the research pipeline.

npbuilds 47e9aa8 2.8 KB Updated Sep 17, 2026

File contents

Evidence Gathering — Director

Research splits cleanly into two halves: acquiring evidence and reasoning from it. This director owns the first half — spelunker's Phases 2 and 3. It decomposes the question into atomic claims, triangulates independent sources for each, traces who funded them, works around paywalls before declaring anything unverifiable, and — for generative questions — hands off to the evolutionary agentic-researcher. The output is qualified evidence bundles ready for synthesis.

Routing Table

Child skill	Use when	Output
`claim-decomposer`	The question bundles multiple assertions, hidden assumptions, or compound predictions	Atomic claims with a dependency graph and priority tiers
`source-triangulator`	An atomic claim needs evidence from multiple independent sources	Per-claim evidence bundle + independence verification
`bias-and-funding-tracer`	Source quality alone is insufficient — you need funding, COI, and cross-source independence	Sources enriched with funder/COI + independence flags
`paywall-strategist`	A critical claim's source is paywalled or inaccessible	Open-access mirrors / abstracts, or a `truly_unverifiable` verdict
`agentic-researcher`	The question is generative ("what's the best X?", "how should we design X?")	Ranked candidate solutions via an evolutionary loop

Routing Logic

claim-decomposer first — everything downstream operates per-claim.
For each claim, source-triangulator. Enrich every critical claim's bundle with bias-and-funding-tracer; if effective independence drops below the confidence tier's threshold, flag it for the synthesizer to down-tag.
If a critical claim's source is paywalled, run paywall-strategist before tagging it Unverifiable.
If the question is generative rather than investigative, route to agentic-researcher instead of the claim pipeline and present its brief in spelunker's Phase 6 format.

Scope Boundaries

In scope: decomposition, source acquisition, source qualification, paywall handling, generative candidate construction.
Out of scope: assembling the brief, adversarial disconfirmation, format translation — all owned by synthesis-verification.
Hand back to spelunker with the qualified evidence bundles; the orchestrator drives the Phase 4 synthesis.

---name: evidence-gatheringdescription: Route a research question to the evidence-acquisition skills — claim decomposition, source triangulation, funding/COI tracing, paywall strategy, and the generative agentic-researcher. Activate during spelunker Phases 2–3 (Decompose → Investigate) when a question needs to be broken into claims and each claim backed by independent sources. This director owns the "find and qualify the evidence" half of the research pipeline.---# Evidence Gathering — DirectorResearch splits cleanly into two halves: acquiring evidence and reasoning from it. This directorowns the first half — spelunker's Phases 2 and 3. It decomposes the question into atomic claims,triangulates independent sources for each, traces who funded them, works around paywalls beforedeclaring anything unverifiable, and — for generative questions — hands off to the evolutionaryagentic-researcher. The output is qualified evidence bundles ready for synthesis.## Routing Table| Child skill | Use when | Output ||---|---|---|| `claim-decomposer` | The question bundles multiple assertions, hidden assumptions, or compound predictions | Atomic claims with a dependency graph and priority tiers || `source-triangulator` | An atomic claim needs evidence from multiple independent sources | Per-claim evidence bundle + independence verification || `bias-and-funding-tracer` | Source quality alone is insufficient — you need funding, COI, and cross-source independence | Sources enriched with funder/COI + independence flags || `paywall-strategist` | A critical claim's source is paywalled or inaccessible | Open-access mirrors / abstracts, or a `truly_unverifiable` verdict || `agentic-researcher` | The question is generative ("what's the best X?", "how should we design X?") | Ranked candidate solutions via an evolutionary loop |## Routing Logic1. `claim-decomposer` first — everything downstream operates per-claim.2. For each claim, `source-triangulator`. Enrich every **critical** claim's bundle with `bias-and-funding-tracer`; if effective independence drops below the confidence tier's threshold, flag it for the synthesizer to down-tag.3. If a critical claim's source is paywalled, run `paywall-strategist` *before* tagging it Unverifiable.4. If the question is generative rather than investigative, route to `agentic-researcher` instead of the claim pipeline and present its brief in spelunker's Phase 6 format.## Scope Boundaries- **In scope:** decomposition, source acquisition, source qualification, paywall handling, generative candidate construction.- **Out of scope:** assembling the brief, adversarial disconfirmation, format translation — all owned by `synthesis-verification`.- **Hand back to `spelunker`** with the qualified evidence bundles; the orchestrator drives the Phase 4 synthesis.

Frequently asked questions