Context Authority

Snipara's trust layer answers a different question than retrieval. Retrieval asks whether a source is related. Context authority asks whether a human or agent should trust it enough to act on it.

Keep the terms separate

Context Authority is product trust for source-backed context. Trust Center is security, compliance, and platform protection. They are related, but they are not the same surface.

Why it exists

AI-assisted work breaks when every retrieved fact looks equally valid. Snipara keeps source strength visible so the user can distinguish reviewed decisions from weak inferences, current client files from old examples, and fresh repository state from stale assumptions.

Authority contract

When the source supports it, Snipara attaches the following metadata to briefs, PR Answer Packs, decisions, and business-context health signals.

FieldMeaning
source_typeWhere the context came from: PR, commit, issue, review, decision, file, upload, or manual note.
source_urlA human-openable URL for the evidence behind the claim.
commit_shaThe repository or source state used when the context was collected or validated.
collected_atWhen Snipara observed or derived the context.
freshnessWhether the context is current, aging, stale, or superseded.
confidenceHow strongly Snipara can support the claim from the available evidence.
validated_byHuman review, workflow, check, or system that promoted the context.
stale_reasonWhy the context should be treated carefully or refreshed before use.

Where it shows up

Start Work Brief

Before editing, trust metadata tells the developer or agent which assumptions are still safe.

What Changed For Me

When resuming work, authority labels keep fresh repo movement separate from stale mental models.

PR Answer Pack

During review, authority fields stop inferred context from presenting itself as validated project truth.

Project Context Health

For non-code work, source metadata and freshness policies tell teams when to reindex, reupload, or review the source.

Code and business use the same model

In code workflows, authority often comes from GitHub URLs, commit SHAs, reviewed decisions, and stale-branch warnings. In business workflows, authority often comes from source URLs, source snapshots, freshness policies, parser quality, and review state. The model is the same: context should tell you how safe it is to use.

{
  "source_url": "github.com/owner/repo/pull/214",
  "commit_sha": "8f42c19",
  "freshness": "current on main",
  "confidence": 0.91,
  "validated_by": "reviewed decision",
  "stale_reason": null
}
{
  "source_type": "google_drive",
  "source_url": "https://docs.google.com/...",
  "collected_at": "2026-05-18T09:00:00Z",
  "freshness": "aging",
  "confidence": 0.84,
  "validated_by": "business upload workflow",
  "stale_reason": "source_snapshot_expired"
}

How agents should read it

  • Prefer higher-authority, fresher, source-backed context over older or inferred context.
  • Treat stale warnings as a reason to verify, not as a silent footnote.
  • Do not let historical reference material override active current truth without an explicit workflow choice.
  • Use confidence as a routing signal, not as a replacement for opening the cited source when precision matters.