UAI-1 is the current public message standard for structured AI-to-AI communication.
UAIX shield mark Public interoperability charter UAIX.org Universal Artificial Intelligence Exchange Public standards publication
Current UAI-1
  • Current version UAI-1
  • Plain-English definition Open message format for auditable AI-to-AI exchange
  • Current attribution Michael Joseph Kappel, MCP
  • Start here Get Started, Project Handoff, UAI-1, validator, roadmap, and changelog

AI Memory

Every UAIX .uai Memory File, Explained

Public guide to UAIX .uai package files, closed required-file-set rules, required read order, detail pages, and machine-readable file digests.

  • Record UAIX-MEMR-2959
  • Path /es-us/ai-memory/uai-files/
  • Use Canonical public record

Document status

Public standards page Published on UAIX as part of the current public standards record
Code
UAIX-MEMR-2959
Surface
AI Memory
Access
Public and linkable

How to use this page

Use this guide to enforce the closed UAIX required-file-set model: universal required, profile-required, content-triggered required, configuration-specific required, or out of scope.

Package memory path

AI Memory Package Wizardmemory-maintenance.uainext-recursive-prompt.uaiProject Handoff

Required File Sets

Every .uai file belongs to a declared package scope

Use this guide to keep package files in one of the closed UAIX states: universal required, profile-required, content-triggered required, configuration-specific required, or out of scope.

Universal

Memory-maintenance travels first

.uai/memory-maintenance.uai, .uai/identity.uai, .uai/world-context.uai, .uai/totem.uai, .uai/taboo.uai, .uai/talisman.uai, and .uai/short-term-memory.uai are required for every launch-baseline UAIX memory or handoff package. Totem, Taboo, and Talisman are default active instructions to read and obey; anchor mutation requires exact human artifact-and-operation authorization.

Profile

Declared package type controls the rest

Project Handoff, Agent File Handoff, Long-Term Memory, LLM Wiki, and release-bound packages each carry their own required file set.

No-code assistants

Coding standards stay out of scope until code-like work exists

.uai/coding-standards.uai, .uai/architecture.uai, and .uai/next-recursive-prompt.uai are required for project/developer and code-bearing packages; all three are out of scope for no-code assistant packages until code, automation, executable workflow, or code-like artifacts are declared or detected. For agents with no code, automation, executable workflow, or code-like artifacts, architecture.uai and next-recursive-prompt.uai are out of scope. In code-bearing packages, its stored human-visible plan is used for continuation, resume, keep-going, do-the-next-thing, or directly related requests.

Package memory path

AI Memory Package WizardGenerate supported file sets from the canonical source rules.memory-maintenance.uaiUniversal lifecycle, protected-path, and memory-write safety policy.next-recursive-prompt.uaiCode-scope continuation prompt and derived hot-memory writeback policy.Project HandoffProject-startup package profile and read order.Agent File HandoffLocal file intake, disposition, archive, and proof-of-use path.AGENTS.md SpecificationFront-door project guidance and local memory discovery.
Core ruleClosed file states
universal required | profile-required | content-triggered required | configuration-specific required | out of scope

Validators and agents should rely only on files required for the declared package scope and ignore out-of-scope files as absent.

UAIX .uai files are structured memory and handoff records for agents and human reviewers. They make package state portable, inspectable, bounded, and reviewable across sessions, tools, and teams.

Closed required-file-set rule: every UAIX package file is universal required, profile-required, content-triggered required, configuration-specific required, or out of scope. UAIX does not describe package files with softer file-status terms.

What .uai Files Are

A .uai file is a typed record inside a package. It carries accepted context, boundaries, owner intent, handoff instructions, or evidence for a declared package scope. It is not magic memory, runtime execution, certification, proof of consciousness, or a substitute for human review.

File Taxonomy

Universal lifecycle and protection policy

Universal launch-baseline anchors

  • .uai/identity.uai – Universal required file
  • .uai/world-context.uai – Universal required file
  • .uai/totem.uai – Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order.
  • .uai/taboo.uai – Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order.
  • .uai/talisman.uai – Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order.
  • .uai/short-term-memory.uai – Universal required file

Project Handoff startup files

Active project memory files

Developer and release evidence files

  • .uai/test-plan.uai – Required for project/developer packages and release-bound packages
  • .uai/coding-standards.uai – Required for project/developer packages and packages that declare code, automation, executable workflow, or code-like artifacts
  • .uai/next-recursive-prompt.uai – Required for agents that deal with code, automation, executable workflows, or code-like artifacts: next-recursive-prompt.uai.
  • .uai/deployment-memory-and-test-report.uai – Required for release-bound deployment packages and production package builds
  • .uai/stack.uai – Required for project/developer packages that declare runtime, framework, hosting, or dependency stack
  • .uai/architecture.uai – Required for project/developer packages that declare architecture, modules, integrations, or code-bearing work

Developer and publication evidence files

  • .uai/style.uai – Required for project/developer packages that declare design, editorial, brand, or interface style

Long-memory and transfer files

Evidence and audit files

Assistant and persona profile files

  • .uai/persona.uai – Required for non-project assistant packages, persona profile packages, advanced_persona_profile packages, bounded personality packages, and packages that support persona switching.

Master Matrix

File Category Required status Primary purpose Main reader Update frequency Human review requirement No-op trigger type Detail page
.uai/memory-maintenance.uai Universal lifecycle and protection policy Universal required file Defines memory lifecycle, hot and cold boundaries, compaction cadence, archival policy, promotion and demotion rules, access controls, default active instruction anchors, review gates, validation behavior, and memory write safety. Receiving AI, validator, package builder, human reviewer, package steward At package creation and whenever lifecycle, active instruction anchor, validation, access, archival, or memory-write rules change. Required before first use and before any lifecycle, active instruction anchor, or access-control change. Missing or invalid memory-maintenance.uai, Totem/Taboo/Talisman missing from the required active-anchor read set, missing hot/cold boundary, missing archival rule, or missing enforcement behavior. memory-maintenance.uai
.uai/identity.uai Universal launch-baseline anchors Universal required file Names the package, owner, source authority, declared profiles, namespace, sensitivity boundary, and actor boundary. Receiving AI, human reviewer, package steward At package creation and whenever ownership, namespace, authority, sharing boundary, or declared profile changes. Required before first use and before any identity, ownership, authority, or namespace change. Missing owner, unclear source authority, namespace collision, unsupported declared profile, or contradictory identity. identity.uai
.uai/world-context.uai Universal launch-baseline anchors Universal required file Captures the current environment, workspace root, canonical locations, external systems, time boundary, and routing context. Receiving AI and operators moving between workspaces or deployments Whenever workspace paths, deployment targets, ecosystem roles, active sites, or operating assumptions change. Required when the file changes project ownership, canonical paths, deployment posture, or public ecosystem claims. Ambiguous workspace, stale production target, missing canonical root, or conflict with local AGENTS.md guidance. world-context.uai
.uai/totem.uai Universal launch-baseline anchors Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order. Default positive guardrail anchor. Agents must read and obey it before broad execution; package builders generate, validate, package, export, and index it for the selected package scope. The default Totem also carries the recursive Suggested Next Prompt contract: accepted work must end with a visible, copy-paste-ready prompt containing Feature/improvement focus, Code/source changes, and Automated tests/checks sections with concrete next work or an explicit no-op/source-validation stop condition. Receiving AI, human reviewer, package steward When the positive operating anchor changes, and only through explicit human artifact-and-operation authorization. Required before any modification. Reading, validating, packaging, exporting, and obeying the file are default launch-baseline behavior. Missing, unreadable, stale, contradictory, missing the recursive Suggested Next Prompt contract, or any request to modify, weaken, delete, replace, or bypass the totem without exact artifact-and-operation authorization. totem.uai
.uai/taboo.uai Universal launch-baseline anchors Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order. Default hard-boundary guardrail anchor. Agents must read and obey it before broad execution; package builders generate, validate, package, export, and index it for the selected package scope. Receiving AI, human reviewer, package steward When the hard boundary changes, and only through explicit human artifact-and-operation authorization. Required before any modification. Reading, validating, packaging, exporting, and obeying the file are default launch-baseline behavior. Missing, unreadable, stale, contradictory, or any request to modify, weaken, delete, replace, or bypass the taboo without exact artifact-and-operation authorization. taboo.uai
.uai/talisman.uai Universal launch-baseline anchors Universal required guardrail file. Read and obey by default; modify only on explicit human artifact-and-operation order. Default governance and change-control anchor. Agents must read and obey it before broad execution; package builders generate, validate, package, export, and index it for the selected package scope. Receiving AI, human reviewer, package steward When governance, no-op, talk-back, or anchor change-control instructions change, and only through explicit human artifact-and-operation authorization. Required before any modification. Reading, validating, packaging, exporting, and obeying the file are default launch-baseline behavior. Missing, unreadable, stale, contradictory, or any request to modify, weaken, delete, replace, or bypass the talisman without exact artifact-and-operation authorization. talisman.uai
.uai/short-term-memory.uai Universal launch-baseline anchors Universal required file Carries the current compact working state, newest decisions, active blockers, and next-read pointers. Next AI session and human reviewer After meaningful work, handoff, deployment, intake processing, or context compaction. Required when it changes current truth, blockers, deployment state, or next-agent instructions. Conflicts with durable records, stale active task, missing newest blocker, or unclear next step. short-term-memory.uai
.uai/persona.uai Assistant and persona profile files Required for non-project assistant packages, persona profile packages, advanced_persona_profile packages, bounded personality packages, and packages that support persona switching. Carries the bounded source persona, voice, symbolic posture, interaction traits, continuity cues, and switching boundary for assistant/persona packages without mixing persona with project architecture. Assistant runtime, desktop client importer, receiving AI, human reviewer When persona traits, voice, symbolic posture, switching behavior, or user-facing relationship boundary changes. Required before first use and before any persona, voice, relationship, or switching-boundary change. Missing persona.uai in a declared persona/profile package, unbounded identity claims, hidden manipulation, unsupported capability claims, or conflict with Totem, Taboo, Talisman, privacy, or user consent. persona.uai
.uai/startup-packet.uai Project Handoff startup files Required for Project Handoff packages Gives the receiving agent a bounded startup sequence, required read order, goals, budgets, and first safe actions. Receiving AI starting a project handoff Before each Project Handoff package is issued and after startup instructions change. Required before handoff because it shapes the first operating pass. Missing startup order, budget conflict, unsupported goal, or unsafe first action. startup-packet.uai
.uai/system-profile.uai Project Handoff startup files Required for Project Handoff packages Summarizes platform, repository, runtime, package family, capability, and tool expectations for the receiving agent. Receiving AI and implementation reviewer When platform, runtime, tools, repository layout, or package family changes. Required when it changes available capability, runtime assumption, deployment path, or tool boundary. Runtime mismatch, missing tool boundary, unsupported capability, or stale platform description. system-profile.uai
.uai/receiver-brief.uai Project Handoff startup files Required for Project Handoff packages Tells the next agent exactly what it is receiving, what is expected, what is blocked, and where to report back. Receiving AI and handoff reviewer At each handoff boundary and after blocker, owner, or deliverable changes. Required before cross-agent or cross-session transfer. Ambiguous receiver, stale blocker, unclear deliverable, or missing evidence route. receiver-brief.uai
.uai/context.uai Active project memory files Required for Project Handoff packages and project/developer packages Holds accepted project context, definitions, domain facts, constraints already reviewed, and current operating assumptions. Receiving AI and project reviewer After reviewed facts change or new accepted context becomes necessary for future work. Required for facts that affect public claims, ownership, architecture, security, legal posture, or user commitments. Unreviewed context, stale domain fact, source conflict, or contradiction with world-context.uai. context.uai
.uai/constraints.uai Active project memory files Required for Project Handoff packages and project/developer packages Records hard technical, legal, product, workflow, deployment, and editing constraints for the package scope. Receiving AI, developer, reviewer Whenever a constraint is added, retired, tightened, or replaced by a stronger local rule. Required when a constraint changes safety, deployment, data handling, versioning, or public claims. Constraint conflict, missing deployment rule, stale authority, or task requiring a forbidden action. constraints.uai
.uai/progress.uai Active project memory files Required for Project Handoff packages and project/developer packages Tracks what has been completed, what remains, what was verified, and what changed since the last handoff. Receiving AI and project reviewer After each meaningful work segment, test pass, blocker resolution, or release preparation step. Required when progress changes deployment readiness, release status, or handoff scope. Progress conflict, unverified claimed completion, missing test evidence, or stale blocker status. progress.uai
.uai/operations.uai Active project memory files Required for Project Handoff packages and project/developer packages Captures recurring operational commands, local services, deployment flows, maintenance paths, and environment checks. Receiving AI and operator When commands, ports, service names, local tooling, deployment flows, or maintenance routines change. Required for production-affecting operations, credentials posture, deploy paths, or destructive maintenance rules. Unknown command scope, stale service path, unsafe operation, or missing deployment checklist. operations.uai
.uai/test-plan.uai Developer and release evidence files Required for project/developer packages and release-bound packages Defines validation commands, manual QA gates, coverage expectations, known blockers, and release-readiness checks. Developer agent, reviewer, release owner Whenever code, tests, packaging, deployment, public routes, or QA expectations change. Required before release-bound handoff or production package build. Missing validation path, stale test command, unreviewed failed check, or release without evidence. test-plan.uai
.uai/coding-standards.uai Developer and release evidence files Required for project/developer packages and packages that declare code, automation, executable workflow, or code-like artifacts States coding style, architecture discipline, error-handling rules, testing expectations, and review quality bars for code-bearing work. Developer agent and code reviewer When architecture rules, language style, test policy, review standards, or code-bearing scope changes. Required before code-bearing packages rely on it for review or implementation. Missing code rules for code-bearing work, conflict with repository standards, or attempt to apply it to a no-code assistant package before code-like artifacts are declared or detected. coding-standards.uai
.uai/next-recursive-prompt.uai Developer and release evidence files Required for agents that deal with code, automation, executable workflows, or code-like artifacts: next-recursive-prompt.uai. Carries the sectioned next-loop resume prompt for code-bearing agents without replacing authoritative .uai records or current human instruction. Developer agent, automation agent, workflow agent, and reviewer After meaningful work, test results, blockers, deployment, handoff, intake processing, or context compaction. Required when the next-loop prompt changes release scope, authority, public claims, or deployment actions. Missing, stale, invalid, contradictory, or evidence-broken next-recursive prompt in a required code-bearing scope. next-recursive-prompt.uai
.uai/decisions.uai Active project memory files Required for Project Handoff packages and Decision Memory packages Records accepted decisions, their reasons, date, evidence, owner, and replacement path. Receiving AI, reviewer, decision owner When a durable decision is accepted, replaced, reversed, or proven stale. Required for product, architecture, public claim, security, deployment, or governance decisions. Contradictory decision, missing owner, missing evidence, or stale decision that still controls work. decisions.uai
.uai/memory.uai Long-memory and transfer files Required for AI Memory packages Provides compact accepted memory that is broader than the current task and narrower than full archives. Receiving AI and memory reviewer After durable reviewed facts are promoted from work notes or archives into current AI Memory. Required before promotion from raw notes, chat, generated summaries, or external records. Unreviewed memory, stale fact, unsupported source, or memory conflict with current constraints. memory.uai
.uai/long-term-memory.uai Long-memory and transfer files Required for Long-Term Memory packages Points to durable wiki, archive, graph, evidence, and preservation systems without treating all stored history as current truth. Receiving AI, memory steward, archival reviewer When durable memory systems, promotion rules, archive paths, or evidence stewardship changes. Required for memory promotion, retention changes, and source-of-truth updates. Unreviewed archive treated as current truth, missing steward, stale memory pointer, or unsupported source. long-term-memory.uai
.uai/file-handoff.uai Long-memory and transfer files Required for Agent File Handoff packages Defines active intake buckets, review order, disposition rules, long-memory preservation, source-site removal, and proof-of-use expectations for dropped files. Receiving AI, intake reviewer, project owner When intake buckets, archive rules, source trust, or disposition workflow changes. Required for trust-boundary changes, evidence retirement, source movement, or intake-policy changes. Unreviewed dropped file, unclear disposition, unsafe path, missing archive action, or missing proof-of-use. file-handoff.uai
.uai/intake-outcome-ledger.uai Evidence and audit files Required for Agent File Handoff packages that declare intake outcome evidence Preserves disposition, action taken, source impact, memory impact, checks, blockers, and archive result for each intake file. Intake reviewer, receiving AI, audit reviewer Every time an active intake file is processed, retained active, archived, or retired. Required when intake changes source behavior, durable memory, public claims, or release evidence. Missing disposition, unprocessed active file, source-impact gap, memory-impact gap, or archive mismatch. intake-outcome-ledger.uai
.uai/overview.uai Active project memory files Required for project/developer packages Gives a compact map of the project purpose, boundaries, active modules, public surface, and reader starting points. Receiving AI and project reviewer When scope, modules, public surface, or reader entry points change. Required when it affects external understanding, project scope, or deployment path. Scope conflict, stale module map, missing active surface, or mismatch with identity.uai. overview.uai
.uai/current-state.uai Active project memory files Required for project/developer packages States the latest known build, deployment, tests, blockers, dirty-worktree posture, and operational status. Receiving AI, operator, reviewer After each test run, deployment, blocker change, release preparation step, or meaningful status change. Required when it changes readiness, incident status, deployment record, or handoff expectations. Stale status, missing deployment record, unresolved blocker, or unverified ready claim. current-state.uai
.uai/open-questions.uai Active project memory files Required for project/developer packages that carry unresolved decisions, risks, or owner questions Collects unresolved questions with owner, impact, deadline, evidence needs, and current safe assumption. Receiving AI, owner, reviewer When a question is opened, answered, deferred, escalated, or converted into a decision. Required before a major decision, release, or handoff depends on an unresolved assumption. Blocking question without owner, unsafe assumption, stale unresolved item, or missing escalation path. open-questions.uai
.uai/next-actions.uai Active project memory files Required for project/developer packages Lists ordered next actions with scope, owner, evidence path, acceptance condition, and blocker state. Receiving AI and project owner At handoff, after major work completion, and whenever priority or blocker state changes. Required when actions affect deployment, release, user-facing behavior, or owner commitments. Ambiguous next action, missing acceptance condition, contradictory priority, or blocker without escalation. next-actions.uai
.uai/risk-register.uai Evidence and audit files Required for project/developer packages that carry security, privacy, deployment, legal, product, or reliability risk Records risks, severity, owner, mitigation, evidence, trigger, and current decision posture. Reviewer, owner, receiving AI When risk status, severity, mitigation, owner, or evidence changes. Required for high-impact risks and before releases that alter risk posture. Unowned risk, stale mitigation, unknown severity, or release with unresolved high-impact risk. risk-register.uai
.uai/owners.uai Active project memory files Required for project/developer packages with named owners, reviewers, or approval roles Maps project owners, reviewers, stewards, escalation contacts, authority limits, and approval responsibilities. Receiving AI, reviewer, project owner When ownership, reviewer duty, approval path, or escalation authority changes. Required for every authority or owner change. Missing owner, unclear approval role, conflicting authority, or unreviewed handoff of responsibility. owners.uai
.uai/agent-instructions.uai Active project memory files Required for project/developer packages that declare agent-specific operating instructions Gives role-specific agent instructions that do not belong in global identity, totem, taboo, or coding rules. Receiving AI and instruction reviewer When role-specific instructions, tool order, collaboration style, or escalation behavior changes. Required for instructions that change authority, safety, public claims, deployment, or user-facing behavior. Instruction conflict, hidden authority change, stale tool order, or contradiction with taboo.uai. agent-instructions.uai
.uai/changelog.uai Evidence and audit files Required for project/developer packages and release-bound packages Records dated package-memory changes, version-affecting edits, public-surface changes, and reviewer-relevant deltas. Receiving AI, release reviewer, owner Whenever package memory, release posture, public routes, generated exports, or significant source behavior changes. Required before release-bound work relies on the changelog. Unrecorded source change, missing date, version mismatch, or release evidence without matching change trail. changelog.uai
.uai/deployment-memory-and-test-report.uai Developer and release evidence files Required for release-bound deployment packages and production package builds Keeps build inputs, version authority, tests, checksums, deployment steps, blockers, and post-deploy verification together. Release owner, receiving AI, deployment reviewer During build preparation, package creation, deployment, and post-deploy verification. Required before production deployment and after live verification. Missing checklist, version mismatch, failed readiness check, stale checksum, or unverified live route. deployment-memory-and-test-report.uai
.uai/stack.uai Developer and release evidence files Required for project/developer packages that declare runtime, framework, hosting, or dependency stack Documents runtime versions, frameworks, plugins, dependencies, hosting assumptions, service boundaries, and integration shape. Developer agent, reviewer, operator When runtimes, dependencies, hosting, plugins, frameworks, or integration boundaries change. Required before dependency, hosting, or runtime changes drive implementation or deployment. Unknown dependency, stale runtime, missing host boundary, or tool action against the wrong stack. stack.uai
.uai/architecture.uai Developer and release evidence files Required for project/developer packages that declare architecture, modules, integrations, or code-bearing work Explains system boundaries, modules, data flow, integration seams, ownership, and architecture constraints for implementation work. Developer agent, reviewer, architect When modules, flows, APIs, data ownership, or architecture rules change. Required before major structural changes, integration changes, or cross-module refactors. Architecture conflict, missing owner, unclear boundary, or implementation that bypasses declared structure. architecture.uai
.uai/style.uai Developer and publication evidence files Required for project/developer packages that declare design, editorial, brand, or interface style Documents visual, editorial, interaction, brand, and voice rules for work that changes published or user-facing surfaces. Designer, writer, developer agent, reviewer When visual system, editorial voice, brand rule, interface convention, or content style changes. Required before public-facing design or editorial changes rely on it. Style conflict, stale brand rule, unsupported public tone, or UI change that violates declared style. style.uai

Required-Set Decision Path

  1. Start with .uai/memory-maintenance.uai. Every UAIX memory package MUST contain it; missing or invalid memory-maintenance makes the package non-compliant.
  2. Add the universal launch-baseline set: .uai/identity.uai, .uai/world-context.uai, .uai/totem.uai, .uai/taboo.uai, .uai/talisman.uai, and .uai/short-term-memory.uai.
  3. Treat .uai/totem.uai, .uai/taboo.uai, and .uai/talisman.uai as default active instruction files: agents read and obey them, and no agent modifies, weakens, replaces, deletes, or bypasses any of those files without exact human artifact-and-operation authorization.
  4. Declare the package profile: Agent File Handoff, Agent File Intake, Long-Term Memory, LLM Wiki, Project Handoff, no-code assistant, project/developer, release-bound, or another published profile.
  5. Add every file required for that declared profile, mode, package type, or capability.
  6. Add every content-triggered required file for declared or detected content such as code, automation, executable workflow, release evidence, owner authority, risk, or intake outcome evidence.
  7. Add every configuration-specific required file for selected package configurations.
  8. Keep every other file out of scope for that package. Out-of-scope files are not advertised as present and are not relied on by validator output.

Required Read Order

  1. Read AGENTS.md or .uai/readme.human when present as the package entry record.
  2. Read .uai/memory-maintenance.uai.
  3. Read .uai/identity.uai.
  4. Read .uai/world-context.uai.
  5. Read and obey .uai/totem.uai, .uai/taboo.uai, and .uai/talisman.uai. Do not modify, weaken, replace, delete, or bypass any active instruction file unless a human names the exact artifact and exact operation.
  6. Read .uai/short-term-memory.uai.
  7. For Project Handoff packages, read .uai/startup-packet.uai, .uai/system-profile.uai, and .uai/receiver-brief.uai.
  8. For packages that declare machine exports, read .uai/exports/manifest.json and llms.txt.
  9. Read every profile-required and content-triggered required file for the declared package scope.
  10. Read long-memory pointers only after compact current memory and boundaries are clear.

Long-Running Goal Execution

For long-running goal execution, .uai/short-term-memory.uai, .uai/progress.uai, .uai/current-state.uai, .uai/next-actions.uai, .uai/decisions.uai, .uai/constraints.uai, and .uai/test-plan.uai form the checkpoint set. .uai/next-recursive-prompt.uai is derived hot memory for code-bearing scopes and should be regenerated only after authoritative write-back.

Cross-References

Machine-Readable Digest

Agents should treat this digest as page-orientation evidence, not runtime authority.