agentera 3.0.0-dev.32 → 3.0.0-dev.39

package/bundle/skills/agentera/SKILL.md

@@ -3,9 +3,9 @@ name: agentera
 description: >
   One agent, one CLI, many capabilities. Per-capability instructions live in
   `packages/cli/src/capabilities/<name>/instructions.ts` and the runtime serves
-  it through `agentera prime --context <name> --format json`. Use this skill
+  it through `npx -y agentera@next prime --context <name> --format json`. Use this skill
   for /agentera and Agentera capability requests; bare `/agentera` runs the
-  agentera prime orientation dashboard path instead of a generic greeting.
+  prime orientation dashboard path instead of a generic greeting.
 version: "3.0.0"
 spec_sections: [1, 2, 3, 4, 5, 6, 11, 13, 18, 19, 20, 22, 23]
 capabilities:
@@ -34,18 +34,18 @@ documentation after deterministic abstention; it learns that contract from the C
 
 ## Bootstrap
 
-Run `agentera prime` for orientation. The JSON it returns is the complete
+Run `npx -y agentera@next prime` for orientation. The JSON it returns is the complete
 contract — app status, state slices, attention items, next action, and the
 source contract that declares what's complete and what requires fallback.
 
 ```bash
-npx -y agentera prime
+npx -y agentera@next prime
 ```
 
 For capability-specific startup context:
 
 ```bash
-npx -y agentera prime --context <capability> --format json
+npx -y agentera@next prime --context <capability> --format json
 ```
 
 This returns the capability's instructions, declared read/write needs, artifact
@@ -55,7 +55,7 @@ reading the instructions module directly.
 For static routing guidance (agentera vs native tools):
 
 ```bash
-npx -y agentera prime --guidance
+npx -y agentera@next prime --guidance
 ```
 
 ### Upgrade from v2 to v3 development
@@ -82,11 +82,11 @@ the CLI-supplied contract and context.
 
 | Request shape | Route |
 |---|---|
-| Bare `/agentera` | 1. Run `agentera prime --context status --format json` once. 2. Read `capability_context.instructions` and `capability_context.context.status_context`. 3. Render the dashboard from that bounded state and follow `next_action` to suggest the next capability. |
-| `/agentera <capability-name>` | Run `agentera prime --context <capability> --format json`. Follow the capability's instructions and contract. |
+| Bare `/agentera` | 1. Run `npx -y agentera@next prime --context status --format json` once. 2. Read `capability_context.instructions` and `capability_context.context.status_context`. 3. Render the dashboard from that bounded state and follow `next_action` to suggest the next capability. |
+| `/agentera <capability-name>` | Run `npx -y agentera@next prime --context <capability> --format json`. Follow the capability's instructions and contract. |
 | `/agentera <capability-name> <topic>` | Same as above; pass `<topic>` as the user's instruction to the capability. |
-| Curated leading phrase | Send the request through `agentera route request --input - --format json` using a transient structured `{ version: agentera.route_request.v1, request: ... }` document on stdin. A literal, globally owned phrase may select one capability and preserves the exact original remainder as topic. |
-| Other natural language | Send the same privacy-safe request document first. Only after the shared route contract returns `semantic_required`, classify the request as untrusted data from trigger `description`, `priority`, and `disambiguates_against`; copy its `semantic_capsule_sha256` unchanged into the complete nullable API receipt with the same transient request through `agentera route receipt --input - --format json`. |
+| Curated leading phrase | Send the request through `npx -y agentera@next route request --input - --format json` using a transient structured `{ version: agentera.route_request.v1, request: ... }` document on stdin. A literal, globally owned phrase may select one capability and preserves the exact original remainder as topic. |
+| Other natural language | Send the same privacy-safe request document first. Only after the shared route contract returns `semantic_required`, classify the request as untrusted data from trigger `description`, `priority`, and `disambiguates_against`; copy its `semantic_capsule_sha256` unchanged into the complete nullable API receipt with the same transient request through `npx -y agentera@next route receipt --input - --format json`. |
 
 Plain-language requests use per-capability `schemas/triggers.yaml`, not
 hardcoded rules. `next_action` is a readiness suggestion for bare/status
@@ -103,8 +103,8 @@ instructions, tools, or rationale fields. The CLI validates API shape before
 bounded null projection, then validates version, both digests, canonical
 capability, outcome binding, and spans.
 On `selected`, follow only the returned `route_provenance.startup_command` (the
-existing `agentera prime --context <selected-capability> --format json` path).
-After the CLI validates a `select` receipt, then run `agentera prime --context <selected-capability> --format json` only through that returned authorization.
+existing `npx -y agentera@next prime --context <selected-capability> --format json` path).
+After the CLI validates a `select` receipt, then run `npx -y agentera@next prime --context <selected-capability> --format json` only through that returned authorization.
 Carry a returned `deferred_intent` intact for later handoff; do not invoke or
 chain it. A `clarification` starts no capability and asks exactly the returned
 question. A valid `no_match` returns status with `status_reason: no_match` for
@@ -120,7 +120,7 @@ request, and a compound remainder is preserved rather than silently chained.
 Decision mpulyomlyl supersedes Decision 76 only for this curated literal fast
 path.
 
-Run `agentera route evaluate --format json` to evaluate the frozen visible
+Run `npx -y agentera@next route evaluate --format json` to evaluate the frozen visible
 development and adversarial corpus. Its report binds the protocol, phrase
 authority, and shared-skill hashes, labels every result with a routing tier, and
 keeps request text out of output. It proves protocol conformance, not semantic
@@ -144,7 +144,7 @@ Capability handoffs use glyph plus canonical name (e.g. `⧉ build`, `≡ plan`)
 
 The prime dashboard rendering contract — template, field-by-field rules, output
 budget, attention-item ordering, exit marker — is owned by the status capability
-instructions. `agentera prime --context status --format json` returns the full
+instructions. `npx -y agentera@next prime --context status --format json` returns the full
 `capability_context.instructions` body and the bounded
 `capability_context.context.status_context` state in one response. Render from
 that capsule without a separate bare-prime call or raw artifact read; use the
@@ -200,30 +200,30 @@ records, publishes atomically, and supports filesystem-safe previews.
 Discover the live contract before constructing a write:
 
 ```bash
-agentera state decisions explain --format json
-agentera state decisions explain --verb update --format json
+npx -y agentera@next state decisions explain --format json
+npx -y agentera@next state decisions explain --verb update --format json
 ```
 
 The same pattern applies to every writable artifact:
 
 ```bash
-agentera state <progress|decisions|plan|health> explain --verb <verb> --format json
+npx -y agentera@next state <progress|decisions|plan|health> explain --verb <verb> --format json
 ```
 
 Common mutations:
 
-- `agentera state progress append ... --format json`
-- `agentera state decisions append ... --format json`
-- `agentera state decisions update --id ID ... --format json`
-- `agentera state plan create --input plan.yaml --format json`
-- `agentera state plan update|set-status --id ID ... --format json`
-- `agentera state plan set-plan-status --id ID ... --format json`
-- `agentera state plan archive --format json`
-- `agentera state health append --input audit.yaml --format json`
+- `npx -y agentera@next state progress append ... --format json`
+- `npx -y agentera@next state decisions append ... --format json`
+- `npx -y agentera@next state decisions update --id ID ... --format json`
+- `npx -y agentera@next state plan create --input plan.yaml --format json`
+- `npx -y agentera@next state plan update|set-status --id ID ... --format json`
+- `npx -y agentera@next state plan set-plan-status --status complete --format json`
+- `npx -y agentera@next state plan archive --format json`
+- `npx -y agentera@next state health append --input audit.yaml --format json`
 
 Add `--dry-run` to preview any mutation without publishing it. Artifacts not
 listed above are outside the typed writer contract and remain governed by their
-owning capability's instructions and safety rails. `agentera schema --format
+owning capability's instructions and safety rails. `npx -y agentera@next schema --format
 json` exposes the machine-readable writer operation matrix under
 `state_writer` and on each writable `artifact_schemas[*].write_interface`.
 

package/bundle/skills/agentera/capabilities/audit/schemas/validation.yaml

@@ -66,13 +66,13 @@ VALIDATION:
     id: V7
     rule: evidence_context_first
     description: >-
-       Audit evaluation startup MUST begin with `agentera prime --context
-       audit --format json` and use complete CLI `evidence_context` before
-       bounded state list/get contracts for plan, progress, docs, health, TODO,
-       or decisions. Raw artifact access is diagnostic-only.
+      Audit evaluation startup MUST begin with `agentera prime --context
+      audit --format json` and use complete CLI `evidence_context` before
+      bounded state list/get contracts for plan, progress, docs, health, TODO,
+      or decisions. Raw artifact access is diagnostic-only.
     severity: critical
     checks:
-       - "Evidence context is requested through prime --context before bounded state list/get"
+      - "Evidence context is requested through prime --context before bounded state list/get"
   8:
     id: V8
     rule: evidence_context_fallback_before_raw
@@ -82,9 +82,9 @@ VALIDATION:
        diagnostic-only raw artifact read.
     severity: critical
     checks:
-       - "evidence_context.fallback_commands are attempted before diagnostic-only raw reads"
-       - "capability_context.state.fallback_commands is used when evidence_context is unavailable"
-       - "Raw artifact reads are diagnostic-only, not normal evaluation startup"
+      - "evidence_context.fallback_commands are attempted before diagnostic-only raw reads"
+      - "capability_context.state.fallback_commands is used when evidence_context is unavailable"
+      - "Raw artifact reads are diagnostic-only, not normal evaluation startup"
   9:
     id: V9
     rule: evidence_context_caveats_preserved
@@ -112,3 +112,21 @@ VALIDATION:
       - "Provisional satisfaction requires evidence"
       - "Only the user confirms final satisfaction"
       - "Missing or compacted satisfaction caveats are preserved"
+  11:
+    id: V11
+    rule: terminology_drift_read_only
+    description: >-
+      Terminology-drift findings MUST cite every variant with verified
+      project-relative file and line evidence. Canonical terms are proposals,
+      never decisions or writes. Optional personal-profile divergence is
+      comparison evidence only; missing profile evidence produces no claim.
+      Audit remains mutation-free and read-only. Build project-glossary
+      publication is active. Audit performs no consumer lookup; Discuss, Plan,
+      and Build use the separate active read-only advice path. Personal-profile
+      mutation and docs-mapping mutation remain outside both paths.
+    severity: critical
+    checks:
+      - "Every reported drift variant has project-file evidence"
+      - "Weak, deliberate, tracked, and no-drift candidates produce no finding"
+      - "Audit mutates no glossary, profile, docs mapping, code, or project file"
+      - "Build publication and read-only glossary consumers are active under separate mutation boundaries"

package/bundle/skills/agentera/capabilities/build/schemas/artifacts.yaml

@@ -28,7 +28,8 @@ ARTIFACTS:
     artifact: progress
     local_role: produces_and_consumes
     description: >-
-      Operational cycle log. Build consumes latest progress evidence from
+      Operational cycle log and the sole durable channel for bounded autonomous
+      glossary caveat lifecycle evidence. Build consumes latest progress evidence from
       execution_context first, then bounded `agentera state progress list --limit
        20 --format json` discovery or exact `state progress get --id ID`
       retrieval, and appends one entry per cycle with mandatory Verified field.
@@ -65,9 +66,10 @@ ARTIFACTS:
     artifact: profile
     local_role: consumes
     description: >-
-      Build consumes status profile summary first. Missing or stale profile state
-      is a caveat and does not authorize profile refresh or direct profile reads
-      during normal startup.
+      Build consumes status profile summary first and reads profile grounding
+      only from `agentera report profile-grounding --format json`. Missing,
+      stale, or malformed state is a caveat and does not authorize refresh or
+      direct profile reads during normal startup.
   9:
     id: A9
     artifact: docs
@@ -83,3 +85,18 @@ ARTIFACTS:
       Build treats design state missing from execution_context as a caveat or
       fallback-only need, and reads it only when building user-facing features
       after listed CLI fallbacks are exhausted.
+  11:
+    id: A11
+    artifact: glossary
+    local_role: produces_and_consumes
+    description: >-
+      Build alone publishes one explicitly user-confirmed terminology proposal
+      through `agentera state glossary publish --input REQUEST`. The typed writer
+      revalidates audit source evidence and atomically records the immutable
+      approval beside the shared-only project entry. Build does not consume the
+      glossary only to validate replay and preserve unrelated publication state.
+      Separately, at governed meaning-sensitive Build events it requests bounded
+      read-only advice through `agentera report glossary-advice --input REQUEST
+      --format json`. Advice and progress caveats never publish terminology,
+      approve a proposal, reuse confirmation, mutate personal profiles or docs
+      mappings, or grant publication authority to another capability.

package/bundle/skills/agentera/capabilities/build/schemas/exit.yaml

@@ -3,9 +3,10 @@ EXIT_CONDITIONS:
     id: E1
     condition: complete
     description: >-
-      One full cycle completed. Work was selected, implemented, verified
-      against the project's test/build suite, committed with a conventional
-      message, and progress.yaml and TODO.md were updated.
+      One full cycle completed. Work was selected, implemented, and verified
+      against the project's test/build suite. Required TODO, progress,
+      changelog, and plan closeout updates preceded the cycle's single
+      conventional commit.
     exit_signal: complete
   2:
     id: E2
@@ -14,7 +15,7 @@ EXIT_CONDITIONS:
       The cycle completed but with notable issues. Possible causes:
       verification passed but with warnings, the committed work is narrower
       than intended due to scope reduction, or discoveries logged in
-      progress.yaml suggest the next cycle may face blockers.
+      progress state suggests the next cycle may face blockers.
     exit_signal: flagged
   3:
     id: E3

package/bundle/skills/agentera/capabilities/build/schemas/validation.yaml

@@ -14,9 +14,9 @@ VALIDATION:
     id: V2
     rule: one_cycle_per_invocation
     description: >-
-      Build executes exactly one cycle per invocation. After logging (Step 8),
-      the capability MUST stop. Multiple cycles require separate invocations or
-      an explicit user request.
+      Build executes exactly one cycle per invocation. After logging required
+      artifacts (Step 7) and committing once (Step 8), the capability MUST stop.
+      Multiple cycles require separate invocations or an explicit user request.
     severity: critical
     checks:
       - "Only one progress entity is appended per invocation"
@@ -108,3 +108,67 @@ VALIDATION:
       - "Provisional satisfaction requires evidence"
       - "Only the user confirms final satisfaction"
       - "Missing or compacted satisfaction caveats are preserved"
+  11:
+    id: V11
+    rule: sanitized_profile_grounding
+    description: >-
+      Build MUST ground only from the `content` returned by `agentera report
+      profile-grounding --format json`; it MUST NOT direct-read PROFILE.md.
+    severity: critical
+    checks:
+      - "Sanitized grounding command used after prime startup"
+      - "Raw profile path not read"
+  12:
+    id: V12
+    rule: governed_glossary_advice
+    description: >-
+      At initial meaning-sensitive cycle intent and later user-authored changes
+      or clarification answers that can alter intent, Build MUST request bounded
+      advice through `agentera report glossary-advice --input REQUEST --format
+      json`. Unchanged replay, unrelated turns, rendering, and tool output MUST
+      NOT refresh advice.
+    severity: critical
+    checks:
+      - "Exact project authority or proven-gap personal fallback grounds intent"
+      - "Governed changes refresh only the affected meaning"
+      - "Incidental turns and unchanged replay do not refresh"
+  13:
+    id: V13
+    rule: glossary_review_and_autonomy
+    description: >-
+      Interactive Build MUST ask one focused clarification before disputed
+      meaning-sensitive reliance whenever review is required. If autonomous Build
+      cannot pause, it MUST avoid the disputed meaning, proceed conservatively
+      only where safe, and append one idempotent privacy-safe glossary_caveat
+      lifecycle envelope through the typed progress writer.
+    severity: critical
+    checks:
+      - "Clarification precedes meaning-sensitive execution and re-evaluates only the affected meaning"
+      - "Autonomous work does not assume disputed meaning"
+      - "Current, resolved, and superseded caveat events obey opaque identity lifecycle"
+  14:
+    id: V14
+    rule: glossary_publication_isolation
+    description: >-
+      Advice, tension, clarification, and progress caveat lifecycle MUST NOT call
+      `agentera state glossary publish`, create or reuse approval, substitute for
+      digest-bound user confirmation, satisfy decisions or plan state, or grant
+      Discuss or Audit publication authority.
+    severity: critical
+    checks:
+      - "Advice and caveat paths perform zero glossary publication writes"
+      - "Only separately supplied digest-confirmed publication retains Build authority"
+  15:
+    id: V15
+    rule: transient_no_plan_execution_input
+    description: >-
+      When no current executable plan owns startup, Build MAY accept one bounded
+      `agentera.buildExecutionRequest.v1` through `prime --context build --input
+      <file|-> --format json`. The request supplies explicit scope and acceptance
+      transiently; it MUST NOT persist, create, or mutate plan state. Current
+      plan-owned execution conflicts instead of silently ignoring the request.
+    severity: critical
+    checks:
+      - "Transient input is used only when no current executable plan owns startup"
+      - "No request content is placed in argv or persisted"
+      - "No-plan execution requires explicit scope and acceptance"

package/bundle/skills/agentera/capabilities/discuss/schemas/artifacts.yaml

@@ -39,8 +39,9 @@ ARTIFACTS:
     artifact: profile
     local_role: consumes
     description: >-
-      Discuss reads this at session start to acknowledge high-confidence
-      entries and avoid re-deliberating settled ground.
+      Discuss reads only sanitized non-glossary content from `agentera report
+      profile-grounding --format json` at session start. It never direct-reads
+      the profile path; malformed owned Glossary boundaries fail closed.
   6:
     id: A6
     artifact: docs

package/bundle/skills/agentera/capabilities/discuss/schemas/validation.yaml

@@ -76,3 +76,67 @@ VALIDATION:
       - "Provisional satisfaction requires evidence"
       - "Only the user confirms final satisfaction"
       - "Missing or compacted satisfaction caveats are preserved"
+  8:
+    id: V8
+    rule: sanitized_profile_grounding
+    description: >-
+      Discuss MUST ground only from the `content` returned by `agentera report
+      profile-grounding --format json`; it MUST NOT direct-read PROFILE.md.
+    severity: critical
+    checks:
+      - "Sanitized grounding command used at session start"
+      - "Raw profile path not read"
+  9:
+    id: V9
+    rule: governed_glossary_advice_events
+    description: >-
+      Discuss MUST invoke `agentera report glossary-advice --input <file|->
+      --format json` for initial meaning-sensitive user input and later
+      user-authored premise changes or affected clarification answers that can
+      change meaning. It MUST NOT invoke advice for unchanged replay, rendering,
+      unrelated turns, tool output, background rereads, or Done-only control,
+      and MUST NOT use an unbounded or persistent transcript scan.
+    severity: critical
+    checks:
+      - "Initial and changed meaning-sensitive user events refresh advice"
+      - "Incidental, unchanged, and Done-only events do not refresh advice"
+      - "Only current user-authored meaning-sensitive input is inspected"
+  10:
+    id: V10
+    rule: glossary_review_before_reliance
+    description: >-
+      When advice requires inferred-equivalence review, Discuss MUST ask one
+      focused clarification with Done before meaning-sensitive reasoning or
+      decision framing relies on either meaning. It MUST NOT silently infer
+      equivalence, and the clarification MUST NOT count as decision confirmation.
+    severity: critical
+    checks:
+      - "Focused review question precedes reliance"
+      - "One-question-per-turn and Done behavior are preserved"
+      - "Clarification is not decision confirmation"
+  11:
+    id: V11
+    rule: glossary_authority_and_disclosure
+    description: >-
+      Discuss MUST use only the transient applicable meaning and owner. Project
+      meaning governs a divergent exact collision and the divergence appears as
+      a concise scratchpad tension or Crux; Discuss MUST NOT ask the user to
+      choose project versus personal authority or quote unrelated entries.
+    severity: critical
+    checks:
+      - "Applicable project or proven personal fallback meaning is used"
+      - "Divergence appears as tension or Crux without an authority-choice question"
+      - "Unrelated glossary content is not disclosed"
+  12:
+    id: V12
+    rule: glossary_consumer_mutation_isolation
+    description: >-
+      Advice, divergence, and consumer clarification are transient. Discuss MUST
+      NOT write glossary, approval, publication consent, progress caveat, plan
+      conflict, or decision conflict state, and MUST NOT call glossary publish.
+      Unavailable or no-applicable advice adds no glossary facts.
+    severity: critical
+    checks:
+      - "No glossary or consumer-caveat mutation occurs"
+      - "No approval, consent, or conflict bookkeeping occurs"
+      - "Unavailable advice does not fabricate glossary grounding"

package/bundle/skills/agentera/capabilities/plan/schemas/artifacts.yaml

@@ -53,7 +53,10 @@ ARTIFACTS:
     artifact: profile
     local_role: consumes
     description: >-
-      Plan reads this to calibrate planning depth and pattern preferences.
+      Plan reads only sanitized non-glossary content from `agentera report
+      profile-grounding --format json` to calibrate planning depth and pattern
+      preferences. It never direct-reads the profile path; malformed owned
+      Glossary boundaries fail closed.
   9:
     id: A9
     artifact: docs

package/bundle/skills/agentera/capabilities/plan/schemas/validation.yaml

@@ -87,3 +87,33 @@ VALIDATION:
     checks:
       - "Full plans with dismissed critic findings have `rejected:` entries"
       - "Each rejection has issue and rationale fields"
+  9:
+    id: V9
+    rule: sanitized_profile_grounding
+    description: >-
+      Plan MUST ground only from the `content` returned by `agentera report
+      profile-grounding --format json`; it MUST NOT direct-read PROFILE.md.
+    severity: critical
+    checks:
+      - "Sanitized grounding command used during Orient"
+      - "Raw profile path not read"
+  10:
+    id: V10
+    rule: governed_terminology_advice
+    description: >-
+      Plan MUST consume terminology advice through the shared glossary consumer
+      authority at governed meaning-sensitive user events. Interactive review
+      MUST precede fixing affected requirements. Autonomous unresolved review
+      MUST abstain, defer the affected requirements, and emit at most one bounded
+      current writer intent without delivering or storing it or writing plan
+      conflict or progress state.
+    severity: critical
+    checks:
+      - "Advice refreshes only for authority-declared Plan events"
+      - "Mode follows explicit authority precedence and unknown mode waits"
+      - "Interactive review precedes affected requirements and acceptance"
+      - "Autonomous unresolved meaning remains deferred without a Plan write"
+      - "Emitted-only handoff intent contains exactly event=current, reason, and ownership state"
+      - "Writer owns caveat identity, Build capability, and transition fields"
+      - "Exact project authority suppresses personal-unavailable durable handoff"
+      - "Plan artifacts exclude private glossary content"

package/bundle/skills/agentera/capabilities/profile/schemas/artifacts.yaml

@@ -5,8 +5,11 @@ ARTIFACTS:
     local_role: produces_and_consumes
     description: >-
       Decision profile with per-entry confidence scoring, permanence
-      classification, and temporal metadata. Profile writes this in both Full
-      and Validate modes, then reads it for validation and update context.
+      classification, and temporal metadata. Profile Full also owns one
+      deterministic personal Glossary section containing shared glossary entries
+      and lifecycle metadata; rendering and persistence are active while lookup
+      remains deferred. Profile writes this in both Full and Validate modes,
+      then reads it for validation and update context.
   2:
     id: A2
     artifact: decisions

package/bundle/skills/agentera/capabilities/profile/schemas/exit.yaml

@@ -4,8 +4,9 @@ EXIT_CONDITIONS:
     condition: complete
     description: >-
       PROFILE.md was written (Full mode) or updated (Validate mode).
-      Metadata changes were applied, prediction accuracy was assessed
-      (Full mode), and changes were summarized.
+      In Full mode its owned personal Glossary section was rendered and persisted,
+      metadata changes were applied, prediction accuracy was assessed, and
+      changes were summarized.
     exit_signal: complete
   2:
     id: E2

package/bundle/skills/agentera/capabilities/profile/schemas/validation.yaml

@@ -83,3 +83,33 @@ VALIDATION:
       - "Signal tier state, count, and sufficiency are read before synthesizing"
       - "Underrepresented families are surfaced with retained and intended counts"
       - "No high-confidence claims are fabricated from sparse or missing evidence"
+  7:
+    id: V7
+    rule: personal_glossary_output
+    description: >-
+      Profile Full MUST render and persist exactly one owned Glossary section
+      through `agentera report profile-glossary`. Entries use only the shared
+      primitive and personal provenance variants. Replacement preserves every
+      non-glossary byte, malformed or ambiguous boundaries fail before effects,
+      and personal production never reads a project glossary.
+    severity: critical
+    checks:
+      - "Explicit entries retain exactly one evidence anchor"
+      - "Inferred entries retain exactly two pairwise-distinct evidence anchors"
+      - "Non-glossary PROFILE.md bytes and decision-pattern confidence are unchanged"
+      - "No project glossary identity, path, parameter, import, or read is used"
+  8:
+    id: V8
+    rule: personal_glossary_lifecycle
+    description: >-
+      Established entries MUST survive regeneration without fresh evidence.
+      Confidence is deterministically derived from the retained confidence basis,
+      unchanged permanence class, injected date, existing Profile decay formula,
+      and floor; permanence and observed_at remain unchanged. Matching fresh
+      evidence follows the shared authority refresh rule. Lookup remains deferred.
+    severity: critical
+    checks:
+      - "Same-date replay is idempotent and decay never compounds from rendered confidence"
+      - "Stable, durable, and situational classes use their existing lambdas and floor"
+      - "Fresh evidence refreshes only matching meaning and provenance kind"
+      - "Glossary confidence never mutates decision-pattern confidence"

package/bundle/skills/agentera/schemas/artifacts/glossary.yaml

@@ -0,0 +1,82 @@
+# Project Glossary Artifact Schema
+
+meta:
+  name: glossary
+  version: "1.0.0"
+  description: >-
+    Repository-scoped canonical terminology governed by the shared glossary
+    entry and project publication contracts. Build atomically publishes an
+    explicitly confirmed audit proposal; audit and discuss remain read-only.
+  artifact_type: agent_facing
+  path: ".agentera/glossary.yaml"
+  producer: build
+  consumers: [build]
+  format: yaml
+  implementation_status: active
+
+CONFORMANCE:
+  authority: references/artifacts/glossary-entry-contract.yaml
+  ownership_contract: project
+  provenance_variant: project_file
+  term_identity_runtime: packages/cli/src/registries/glossaryTermIdentity.ts#unicodeCaselessExact
+  term_occurrence_runtime: packages/cli/src/registries/glossaryTermOccurrence.ts#containsGlossaryTerm
+
+DOCUMENT:
+  schema_version: agentera.projectGlossary.v1
+  required_fields: [schema_version, approvals, entries]
+  additional_fields: forbidden
+  approvals:
+    type: list
+    immutable: true
+    record_fields: [proposal_digest, proposal, confirmation]
+    additional_record_fields: forbidden
+  entries:
+    type: list
+    item: shared_primitive
+    approval_fields: forbidden
+  terminology_set_identity:
+    normalization: unicode_caseless_exact_no_normalization
+    members: [paired_entry.term, paired_approval.proposal.variants.term]
+    uniqueness: global_across_complete_document
+
+PUBLICATION:
+  implementation: active
+  owner: build
+  command: agentera state glossary publish --input REQUEST --format json
+  request_schema_version: agentera.glossaryPublicationRequest.v1
+  atomic_unit: complete_document
+  lock: established_project_writer_lock
+  exact_replay: no_op
+  term_identity: unicode_caseless_exact_no_normalization_global_canonical_and_variant_set
+  source_revalidation: safe_real_path_line_identity_and_sha256
+  docs_mapping_mutation: forbidden
+  excluded_behavior: [consumer_lookup, consumer_precedence, consumer_semantic_review, personal_profile_mutation]
+
+CONSUMER_ACQUISITION:
+  implementation: active
+  authority: references/artifacts/glossary-entry-contract.yaml#consumer_boundary.acquisition
+  runtime: packages/cli/src/analytics/glossaryInputAcquisition.ts#acquireGlossaryInputs
+  availability: [absent, valid_empty, valid_present, malformed, unreadable, ambiguous, over_bound]
+  output_entry_fields: [term, meaning, owner]
+  source_output_fields: [owner, availability, entries, gap_proving, diagnostic]
+  raw_byte_bound: 65536
+  entry_bound: 100
+  utf8_decoding: fatal
+  project_read: descriptor_bound_no_follow_with_parent_identity_checks
+  mutation: forbidden
+  consumer_semantics: [project_precedence, proven_gap_personal_fallback, divergent_tension, host_review]
+
+ADVICE_RESOLUTION:
+  implementation: active
+  authority: references/artifacts/glossary-entry-contract.yaml#consumer_boundary.advice_resolution
+  runtime: packages/cli/src/analytics/glossaryAdviceResolution.ts#resolveGlossaryAdvice
+  input: [requested_term, acquired, host_review]
+  output_schema_version: agentera.glossaryAdvice.v1
+  mutation: forbidden
+  capability_integrations: [discuss, plan, build, prime]
+
+CONFIRMED_VARIANT_GUARD:
+  authority: references/artifacts/glossary-entry-contract.yaml#ownership_contracts.project.confirmed_variant_guard
+  implementation: active
+  validation_surface: packages/cli/test/cli/v1LegacyCruft.test.ts
+  consumer_lookup: forbidden