- Home
- Skills
- Research & Analysis
- Research Decision Skill — Evidence-Based STEM Decisions
Research Decision Skill — Evidence-Based STEM Decisions
Choose research directions, compare ideas, assess evidence, and decide which experiment to run next.
Free
Works with the AI Agents you already use
About this skill
Make evidence-backed research decisions when choosing a research direction, comparing alternatives, evaluating an idea, or deciding what experiment to run next.
Version — v0.7.0-preview
Agensi platform revision v1.1 was approved on 22 October 2026 and uses the unmodified official GitHub v0.7.0-preview release (9 October 2026). Agensi revision numbers and upstream release numbers are separate. The uploaded generic ZIP contains 68 files, including 66 core Skill files, EXPORT.json and START-HERE.md.
Use it when you need to
- Choose a research question and narrow a claim to something testable.
- Compare methods or competing hypotheses against your goals, evidence, and constraints.
- Design the next experiment, proof, or discriminating test.
- Interpret failed, anomalous, or unstable results and decide whether to continue, pivot, or stop.
What you get
A concise decision card: a conditional recommendation, the key reasons and source links, one next action, and the evidence that would change the recommendation. Ask for a deeper explanation, an alternatives comparison, or an audit record when useful.
The skill uses a nonlinear map of 29 research decisions, selects the relevant STEM discipline pack, and searches for current public expert reasoning. It checks how sources apply to your situation and can defer judgment when evidence is insufficient. Project files can preserve the reasoning and evidence across research sessions.
What changes in upstream v0.7
Building on v0.6, v0.7 keeps follow-up explanations and new results connected to the existing decision, considers equipment, samples, costs and scheduling, adds 8 method cards and attributed research-practice adapters, and provides optional context/artifact binding. GitHub includes 16 synthetic behavior cases and six installation layouts: Claude Code, Codex, Copilot, Cursor, Gemini CLI and generic. These layouts share the same core files; layout checks do not establish cross-agent performance.
Who it is for
Scientists, engineers, and STEM students working in AI/ML, mathematics and statistics, computer science, physics and astronomy, chemistry and materials, engineering, life sciences research, earth and environmental sciences, or interdisciplinary STEM research.
Getting started
On Agensi, use the Agensi MCP connection to load the latest approved platform version in your agent. Agensi currently delivers skills through the agent connection rather than a buyer ZIP download.
For a complete local installation, download the matching host ZIP or generic ZIP from the official v0.7 release, check SHA256SUMS.txt, and follow START-HERE.md and the installation guide. Keep the complete skill directory, including references, scripts, schemas and templates. Preserve your old installation and research records when upgrading; do not mix files from different releases.
Share your research goal, work completed, current results, unresolved decision and constraints. Ordinary skill use does not require Python. Optional local record and artifact tools require Python 3.10+ and authorized file access. Search, file access, helper execution and persistence depend on the host. Full Agensi MCP runtime behavior has not been independently verified.
Upstream v0.7 retains schema v3 and reads same-schema v0.3–v0.6 records without rewriting them. Legacy schema v1/v2 remains read-only. Read the changelog.
Watch the 66-second product introduction. Its scenarios and operations are scripted illustrations, not real experimental results.
Free, open source, and attributed
Created by Odyphus. This listing remains free. GitHub is the canonical source for code, documentation, updates, and issues: github.com/Odyphus/research-decision-skill.
Original code and wording are licensed under Apache-2.0. The submitted official ZIP preserves LICENSE, NOTICE, CITATION.cff, LICENSE_AND_ATTRIBUTION.md, THIRD_PARTY_NOTICES.md and the attributed research-practice licence notices. The separate promotional video source uses AGPL-3.0 and is not part of the Skill package. The decision map draws on Price et al. (2021), DOI 10.1187/cbe.20-12-0276. Their article and supplement are not bundled or relicensed; this is an independent product adaptation.
This is a public-source decision-support Preview. Human domain-expert validation and durable learning-transfer effectiveness have not been demonstrated.
Public-source STEM research decision support, still Preview. Agensi platform v1.1 corresponds to upstream v0.7.0-preview. Human domain-expert validation, durable learning-transfer effectiveness and broad cross-agent performance are not demonstrated. Agensi MCP access to all supporting files, helpers and persistent project records has not been independently verified. Source quality, tool access and project evidence limit recommendations. No clinical diagnosis or treatment; no unsupported non-STEM expert judgments. Legacy schema v1/v2 records remain read-only.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
This skill has been used 8 times.
Trust & safety
Security scan
Security scan passed 27 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes
Public web search/network access is used to retrieve and verify current sources. Reading project evidence requires user-authorized file access. Optional local record and artifact helpers require Python 3.10+, shell access and authorized local writes; ordinary discussion does not require these helpers. Research state stays in the project .research-decision/ directory, or supported legacy .research-exploration/ when used alone. No API key or environment-variable harvesting is required by the Skill. Agensi MCP availability of supporting files and helpers has not been independently verified.