skills.rest
pedrohcgs
pedrohcgsCOMMUNITY
Explore all skills by this author →
Website𝕏@pedrohcgsGitHub

credible-claims

Enforces research briefs and claim records for delegated AI-assisted research work.

1.5k|3.0k|Updated Feb 6, 2026
Search Tags:#verification#delegation#evidence#reproducibility#research brief#Academic research#claim record
One-click install
npx skills add https://github.com/pedrohcgs/claude-code-my-workflow --skill credible-claims
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: credible-claims
Source: https://github.com/pedrohcgs/claude-code-my-workflow/tree/main/.claude/skills/credible-claims
Command: npx skills add https://github.com/pedrohcgs/claude-code-my-workflow --skill credible-claims

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Delegated or AI-assisted research often produces confident conclusions without inspectable evidence, making it hard to know whether results actually support the claims in a paper or decision. This Skill imposes a lightweight discipline: a research brief written before execution and a claim record produced during or after execution.

Core Features & Use Cases

  • Research Brief: Defines the question, completion criteria, prohibited substitutions, known failure modes, required evidence, and escalation triggers before any delegated run starts.
  • Claim Record: Documents what supports each claim, what changed after seeing results, what remains unresolved, and who decided what could be reported.
  • Reporting Language Discipline: Prevents upgrading language beyond the evidence (associational vs causal, pointwise vs uniform, illustrated vs validated).
  • Use Case: Before launching an autonomous agent to run simulations for an econometrics paper, write the brief specifying the estimand and prohibited substitutions; afterward, produce a claim record distinguishing diagnostic-triggered fixes from favorable switches.

Quick Start

Ask the AI to write a research brief before starting the analysis task and a claim record when reporting the results.

Frequently Asked Questions about credible-claims

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep AI agents from overclaiming research results?▼

Write a research brief before delegating that defines the question, completion criteria, and prohibited substitutions, then require a claim record after execution documenting evidence, changes made after seeing results, and unresolved checks. This separates faster execution from credible evidence.

What is a research brief for delegated AI work?▼

A research brief is a short block written before launching delegated work specifying the question or target, completion criteria, prohibited substitutions, known failure modes, required evidence, escalation triggers, and a blocked-route rule for paths depending on unavailable data or unproved results.

When should I use a claim record in research workflows?▼

Use a claim record whenever results will support a claim in a paper or decision. A routine task needs only one short paragraph; heavier records apply when branching is extensive, outputs will be reused, errors are consequential, or correction is costly.

Does this skill work with any AI agent or only Claude Code?▼

The discipline applies to anything delegated, including agents, workflows, simulations, proof patches, and data construction. It is packaged as a Claude Code skill but the brief-and-record protocol is tool-agnostic.

What decisions must be escalated back to the researcher?▼

Anything that changes the economic object, identifying assumptions, inferential procedure, or reporting language must return to the researcher. A standing user ruling counts as a returned decision and should be recorded in the claim record.

Related Skills

Explore More Education & Research Skills→
fredericpoitevinfredericpoitevin

clarify-before-research

Ask clarifying questions to define research goals and scope before starting.

Community
Basic
goodfire-aigoodfire-ai
108

replicate-paper

Extract methodology and experiment lists from research papers to plan and run replication pipelines.

Official
Advanced
devrelaicomdevrelaicom
37

midnight-verify:verify-sdk

Classify SDK/TypeScript claims and route them to verification methods.

Official
Advanced
K-Dense-AIK-Dense-AI
282

preregistering-analysis

Enforce timestamped preregistration of hypotheses and analysis plans before results.

Official
Advanced