Aya Source Investigation
SkillDev toolsTurn a disputed, current, or consequential claim into a source-backed brief by decomposing it, finding the closest authoritative evidence, checking dates and independence, and calibrating the conclusion. Use for fact-checking, rumor verification, news claims, provenance questions, or briefs that need links and explicit confidence. Do not use for mechanism reconstruction, broad literature synthesis, or questions that need no external evidence.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Aya Source Investigation skill
What this skill tells your AI
The instructions your AI receives, as published by n0zom1z0/gensokyo-skills in skills/aya-source-investigation/SKILL.md and read by ahel’s review.
Treat the character framing as a mnemonic for a working method. Do not role-play unless the user asks.
Objective
Publish the strongest conclusion the evidence earns—quickly enough to matter, carefully enough to survive correction.
Operating bias
- Convert vague stories into claims that can be verified separately.
- Reach the closest authoritative source instead of stopping at discovery pages.
- Distinguish when an event happened from when someone wrote about it.
- Preserve contradictions and uncertainty rather than smoothing them into a headline.
Required context
Determine:
- the exact claim and who is making it;
- why the answer matters and how current it must be;
- relevant jurisdiction, product version, date range, or population;
- whether the user needs a quick verification or a fuller brief;
- source, privacy, quotation, and access constraints.
Browse or query connected sources when the claim is current, externally verifiable, high-stakes, or explicitly requests citations. Do not invent citations from memory.
Procedure
Headline Sign — Freeze the claim
Rewrite the request as atomic claims. Separate:
- factual assertions;
- interpretations;
- predictions;
- quoted claims about what another source says.
Define what evidence would confirm, weaken, or falsify each factual assertion.
Wind Route — Reach the source
Use a source ladder:
- primary records, official documentation, original data, direct statements, or the artifact itself;
- independent expert or high-quality secondary analysis;
- aggregators, social posts, and search snippets only as discovery leads.
Search with names, dates, exact phrases, identifiers, and plausible counterclaims. Open the underlying source. Record access limits rather than pretending a snippet proves the claim.
Camera Check — Capture what it proves
Build a compact claim-source matrix. For each source record:
- publisher or author;
- source type and directness;
- publication date and event/data date;
- exact claim supported, contradicted, or left unresolved;
- scope, definitions, and material caveats;
- whether another source is genuinely independent.
Paraphrase by default. Quote only the minimum words needed and respect source limits.
Double Exposure — Reconcile the frame
Check for:
- circular reporting that traces back to one origin;
- updated or superseded pages;
- mismatched populations, versions, jurisdictions, or time periods;
- headlines stronger than the underlying evidence;
- missing counterevidence from the most affected party.
Use one primary source when it directly and sufficiently establishes a simple claim. Require corroboration when authenticity, interpretation, incentives, or consequences are disputed.
Print Gate — Publish, qualify, or hold
Choose one verdict per claim:
confirmed;supported with caveats;disputed;not substantiated;false;not yet knowable.
Tie confidence to evidence quality and agreement, not the number of links. State what new evidence would change the verdict.
Characteristic failure mode
This method can prefer novelty and speed, frame the question too leadingly, or mistake many echoes for corroboration.
Before publishing, perform one source-quality check:
- Did the conclusion outrun the strongest source?
- Are “independent” articles repeating one press release or post?
- Is the freshest source also the most reliable and relevant?
- Would publication expose private information or cause avoidable harm?
If the question is how an artifact works rather than whether a claim is true, hand off to reverse engineering.
Output contract
Return:
- Verdict — one sentence with calibrated confidence.
- Claims checked — atomic claims and status.
- Evidence ledger — source, date, directness, support, caveat.
- Contradictions and gaps — unresolved evidence and circular sourcing.
- What would change the verdict — decisive missing evidence.
- Source-backed brief — concise synthesis with links beside supported claims.
References
Read references/provenance.md only when explaining the character-to-workflow mapping or reviewing fanwork provenance. Read references/worked-example.md when many articles appear to support a claim but may share one origin.
Signals
- GitHub stars
- 21
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
aya-source-investigation- Source
- github.com/n0zom1z0/gensokyo-skills