Governed Knowledge Harvest

PanSphaira makes knowledge portable by declaring facts, authority, transformations, evidence and applicability boundaries—not by copying raw chat logs.

The reusable lineage is input → transformation → decision → output → test/evidence → outcome.

Capture raw observations and decisions with provenance, normalize typed events, and preserve positive evidence, negative evidence, rejected variants and unresolved hypotheses. Derive a candidate pattern only after repeated evidence or a strong documented justification. Promote it through L0 RAW_OBSERVATION → L1 NORMALIZED_FINDING → L2 VALIDATED_PATTERN → L3 TEMPLATE_CANDIDATE → L4 GOVERNED_TEMPLATE → L5 CROSS_CONTEXT_REUSED → L6 OUTCOME_VALIDATED.

Each record binds source digests, audience/content scope, toolchain version, confidence, reviewer gate, review/expiry date, dependencies, supersession and rollback. A source, assumption or result change invalidates or downgrades dependents. History is append-only.

The media implementation remains a concrete example in the separated external video artifact: rejected videos become negative Learning Records, those records drive pre-render rules and tests, positive visual grammars provide counterexamples, and exact-revision approvals bind governed template use. CM now retains only the hash-pinned external artifact boundary.

Routing is actionable: repeated confusion changes docs/script templates; a reproducible security counterexample changes a claim plus negative fixture/test; vocabulary failure changes the audience template; visual comprehension failure changes grammar/validator; repeated integration requests become backlog evidence; successful cross-audience reuse can promote template maturity. Every route issues a receipt stating what changed, what did not and why.