artifact-analysis

Type: types/tag-readme.md

Artifact analysis classifies retained artifacts by storage substrate, representational form, lineage, and behavioral authority. Assign this tag to work that defines, extends, tests, or draws design consequences from that scheme. Merely storing artifacts or filling in the fields is insufficient. Document-system covers writing and checking documents; this tag covers retained artifacts across forms. A child of learning-theory.

The scheme

• axes of artifact analysis — the four-field analysis: storage substrate, representational form, lineage, and behavioral authority over the operative part or consumption path
• first-principles analysis maps a design space before selecting within it — uses representational form as the worked case for deriving designer options from primitive axes before inherited packages close the search space
• the four-field record exposes an efficiency, security, and sovereignty agenda — what the classification is for: the downstream questions the fields make answerable
• memory design adds operational axes to artifact analysis — extends the scheme with write/read-side operational axes for memory-system design

Definitions

• retained artifact — retained state a later agentic loop consumes in a behavior-shaping way; the boundary is behavioral consequence, not storage label
• operative part — the behavior-affecting content, structure, parameterization, or mechanism within a retained artifact
• storage substrate — where retained state persists; storage is one field, not the taxonomy
• representational form — how an operative part is encoded and consumed; form determines the default review method
• lineage — source dependencies and derivation status needed to invalidate, regenerate, retire, or review retained behavior
• behavioral authority — who consumes an artifact, through which channel, with what force
• knowledge artifact — retained artifact consumed as evidence, reference, context, explanation, or advice
• answerability — the stipulated domain-admission invariant for knowledge: a nameable referent, asserted property, and correction-triggering discrepancy
• system-definition artifact — retained artifact consumed with instruction, enforcement, routing, validation, configuration, evaluation, or learning force

Applications

• an action model matters only through its consumption path — applies operative parts and behavioral authority to the difference between direct and model-mediated agentic action
• system-definition artifacts are crystallized reasoning under context pressure — why system-definition artifacts exist at all: reasoning crystallized so it need not be redone in-context
• Artifact classification separates content kind, lineage, and authority — separates region-level belief or residual choice, production lineage, and path-relative authority; collection contract and proposal status are different whole-artifact surfaces
• superseded choices are retained; superseded beliefs are not — what becomes of displaced content once the maintenance operation is chosen: the retention obligation follows content kind, so append-only chains and holistic rewrite are the two regimes
• an artifact must preserve the scope of each named system choice — each proposition needs a recoverable reference rule or range plus the choice's role within context guaranteed to its consumers
• Local materialization should outperform distant natural-language declarations — statistical conjecture that a generated local view from one canonical fact should improve correct application over declaration-only presentation for distant or non-obvious uses
• adaptation signals choose pressure; artifact analysis chooses the retained surface — applies the four-field record to agentic-adaptation signals, separating update pressure from the retained surface and review evidence it requires
• a consumption channel delivers force without the history that earned it — the failure surface behind path-relative authority: injected content and honest error both act because the channel reads position, not provenance
• six Commonplace paths establish broad addressability, not completeness — a bounded authority-path audit that establishes broad coverage, leaves completeness open, and exposes applicability missing from the current record
• stale self-description conceals its own staleness — what the reflexive case adds to artifact drift: the detecting process consults the artifact that drifted, the trigger has no edit event, and load scales with autonomy
• for its load-bearing part, documentation generates the system rather than describing it — a lineage test runnable now: attempted recovery in both directions, sorting a system's documentation into cache and generator content
• seven documentation cases left routing and synthesis — a bounded Commonplace sweep in which exact source copies collapsed while discovery maps and cross-component relations survived
• verification needs a typed target before it needs an oracle — why classification precedes evaluation: an oracle's domain is a set of artifacts, so an unclassified layer yields verdicts without warrant
• orchestration strategies and run state have opposite persistence — a persistence asymmetry the fields predict: strategies accumulate, run state is disposable
• ephemeral computation prevents accumulation — discarding generated artifacts trades accumulation for simplicity; the inverse of codification
• ephemerality is safe where embedded operational knowledge has low explanatory-reach — the boundary rule: discard is safe below the explanatory-reach threshold, accumulation mandatory above it
• RLM, Tendril, and llm-do place symbolic work at different persistence points — three systems located by where their symbolic artifacts persist