More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Epistemic Provenance: AI Trust & Source Tracer

    1

    Tiers every claim by trust level and flags likely-synthetic content before you publish it.

    $29

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Epistemic Provenance: AI Trust & Source Tracer

    Epistemic Provenance: AI Trust & Source Tracer

    Example session with this skill installed

    I'm building a study guide on the human brain for my students. Can you check these two claims before I include them: "Humans only use 10% of their brain" and "The Wright brothers' first powered flight was December 17, 1903, at Kitty Hawk."

    • Read your context and instructions
    • Compiled the epistemic provenance

    Checked both claims and tiered them — one needs a fix before it goes in your guide.

    [Tier 2 — Widely repeated, and actually FALSE]
    "Humans only use 10% of their brain."
    This is one of the most cited "facts" in pop psychology, but it's been debunked repeatedly. Brain imaging studies show activity across virtually the entire brain over the course of a day — no region sits permanently dormant. The myth likely originated from a misquote of early 20th-century psychologist William James and has been falsely attributed to Einstein ever since. Recommend cutting this from the study guide entirely, or flipping it into a "myth-busting" section instead of stating it as fact.

    [Tier 1 — Verified]
    "The Wright brothers' first powered flight was December 17, 1903, at Kitty Hawk, North Carolina."
    Independently corroborated by primary historical records (Smithsonian, National Park Service archives) that don't trace back to a single origin. Safe to include as-is.

    Summary table for your guide

    ClaimTierAction
    "10% of brain"2 (false)Remove or reframe as a myth
    Wright brothers date1 (verified)Keep as written

    Next steps

    1. Swap the brain-usage claim for an accurate one (e.g., "different brain regions activate for different tasks, not a fixed 10%")
    2. If this guide will be sold or distributed, consider running the finished draft back through this skill once more — new claims sometimes get added during editing
    3. Add a one-line provenance note in your document footer using the included tag template, so future-you knows which facts were checked

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most AI tools state everything in the same confident tone — a verified fact and a lucky guess sound identical. That flatness is the actual risk right now: fabricated citations, unverified stats repeated as truth, synthetic content quietly polluting the training commons.

    Epistemic Provenance fixes this at the output level. Instead of a flat true/false answer, it breaks any input — a claim, article, quote, dataset, or image — into checkable units and assigns each one a confidence tier:

    • Tier 1 — Verified (traceable to a primary or independently-corroborated source)
    • Tier 2 — Widely repeated (common but no single traceable origin)
    • Tier 3 — Single source (one origin, no corroboration — flagged with its incentive)
    • Tier 4 — Unsourced inference (plausible-sounding, but not checked)

    It also screens text and images for synthetic-content signals (reported as probabilistic indicators, never false certainty), and gives creators a lightweight chain-of-custody tagging system — [source], [AI-synthesized], [SME] — so an AI-assisted product's origins stay auditable months later.

    Built for educators, consultants, content creators, and agencies producing anything a client or student will treat as fact: lesson plans, sales copy, research summaries, worksheets, reports.

    Works with any SKILL.md-compatible agent — Claude Code, Cursor, Codex CLI, and more.

    Known Limitations

    • Not a fact-checking oracle — accuracy depends on the host agent having web search. Without it, unverifiable claims default to Tier 4 rather than being guessed at.

    • Not a substitute for domain experts on high-stakes claims (medical, legal, financial, safety-critical) — this triages confidence, it doesn't certify truth.

    • Synthetic-content detection is probabilistic, not forensic. Signals, not proof, and they degrade as generation models improve.

    • No true reverse image search — image provenance is limited to what the agent can search or the user supplies.

    • Tiering quality scales with search quality — offline or search-limited sessions will default more claims to Tier 3/4, even where a verified source exists elsewhere.

    • Doesn't tier opinions or values-based claims — only checkable factual claims.

    • Same claim run twice may occasionally land on adjacent tiers (e.g., Tier 2 vs. 3) depending on what that run's search turns up.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 2 months ago

    Permissions

    No special permissions declared or detected

    Listed2 months ago

    What's inside

    Frequently Asked Questions

    Epistemic Provenance breaks down claims into checkable units and assigns one of four confidence tiers (from Verified to Unsourced Inference), allowing you to see exactly which parts of a response are backed by evidence and which are speculative.

    It is designed for any SKILL.md-compatible environment, including Claude Code, Cursor, and Codex CLI, allowing you to integrate source-tracing directly into your existing development or writing workflow.

    Beyond source-tracing, the skill screens for probabilistic signals of synthetic generation and provides a 'chain-of-custody' tagging system ([source], [AI-synthesized], [SME]) to maintain an audit trail for your content.

    Your purchase includes the core Epistemic Provenance skill logic, the confidence tiering framework, the synthetic-content signal reporter, and the chain-of-custody tagging system.

    Installation is straightforward: you simply import the skill into your SKILL.md-compatible agent's directory, where it begins analyzing outputs based on the defined provenance tiers.

    New in Research & Analysis