Memory and knowledge boundaries

Evidence Lane uses several owner-specific memory surfaces. They cooperate through typed locators and receipts without becoming one generic agent memory.

Current counts are derived release facts, not permanent ceilings.

Surface	Retention and role
Current prompt/steer	Short-lived input classified by Entry Slip and Source Intake
ChatLineage	Visible task prompts, responses, steers, events, and entry/exit boundaries
Project Memory	Durable bounded locators and relationships across project authorities
Agent Learning	Accepted reusable lessons under separate Learning HIL
Canon	Typed cross-task contracts, messages, and result continuity
Project Universe	Per-project relationship graph
Instructions	AGENTS.md and host MEMORY.md chain, nonauthoritative to Project Truth

SQLite/FTS5/BM25 remains the durable bounded retrieval authority. Optional local semantic indexes can rank candidates but cannot replace canonical IDs or owning SQLite. Raw project databases, whole Markdown histories, or full chat archives are not dumped into model context.

PreCompact seals the active continuity boundary; PostCompact rehydrates the same bounded Plan/Goal/task/source coordinates. Compaction is not a new PV, decision, or Goal.

Source-bound workflow map

This page is projected from the same current executable snapshot as the rest of the documentation set. The map is deliberately two-directional: each horizontal district shows peer stages while vertical edges show ownership and state progression.

flowchart TB
    subgraph InputDistrict["Input and classification"]
      direction LR
      A["Visible project or task evidence"] --> B["Owner-specific memory classification"] --> C["ChatLineage, Memory, Learning, Canon, or Universe"]
    end
    subgraph ExecutionDistrict["Selection and execution"]
      direction TB
      D["SQLite and bounded indexes"] --> E["Exact locator query"] --> F["Authority and provenance validation"]
    end
    subgraph EvidenceDistrict["Evidence and outcome"]
      direction LR
      G["Memory query receipt"] --> H["Bounded context projection"]
      G -. mismatch .-> I["Never merge memory authorities"]
    end
    C --> D
    F --> G

Contract and readback

Phase	Current contract	Required readback
Input	Visible project or task evidence	Exact identity, provenance, and scope
Classification	Owner-specific memory classification	Owning schema, action, lane, skill, or authority
Owner	ChatLineage, Memory, Learning, Canon, or Universe	One canonical implementation owner
Route	SQLite and bounded indexes	Condition-true ordered route with no hidden alias
Execution	Exact locator query	Real execution or a visible fail-closed result
Validation	Authority and provenance validation	Hash, schema, authority-effect, and negative-case checks
Receipt	Memory query receipt	Content-addressed result and provenance receipt
Downstream	Bounded context projection	Only the explicitly eligible next state
Failure	Never merge memory authorities	No inferred HIL, candidate acceptance, or pointer movement

Canonical source owners

authorities/project_memory/manifest.v1.json
schemas/memory/project-memory.v1.sql
skills/evi-memory/SKILL.md

Exact backend readback

Source contract	Bytes	SHA-256
`authorities/project_memory/manifest.v1.json`	7792	`B85EF6F68A84C0263945E3F4CFAD222F209203FE31D46D9F5C927A7D5084242A`
`schemas/memory/project-memory.v1.sql`	3281	`4622D826BA67B21589255EB58211FA954C51AA4148FACCFEC2B498F0C491C17B`
`skills/evi-memory/SKILL.md`	2153	`6617AD64BDA4ABF6E4FA5EB62EAA614D5C609DF1332101420F32233181E7DC09`

Cross-surface invariants

The current snapshot contains 91 public actions, 26 skills, 11 hook events / 44 handlers, 119 tool requirements, 18 sector lanes, and 11 named authorities. These are derived counts, not fixed ceilings.
Executable ownership stays one-way: skills select, MCP exposes, the outer SDK routes, the internal SDK executes, ENV selects, UOP governs, tools perform bounded work, hooks emit receipts, and the owning authority validates effects.
Any missing identity, schema, grant, capability, dependency, receipt, or authority proof must fail closed at its owning phase; a later green check cannot retroactively authorize the skipped boundary.
A changed route refreshes every dependent schema, manifest, generator, test, diagram, and documentation reference; the superseded executable route is directly purged in the same Delta.
Tests, Git, CI, installation, restart, deployment, discussion, or a rendered page never imply Project HIL, Learning HIL, Goal completion, or pointer movement.

This page is a Git-tracked documentation projection. Executable source, SQLite authorities, installed-runtime receipts, and explicit human gates remain the governing evidence.