v0.26.10 - pk-doctor false-actionable cleanup

v0.26.10 is a patch release that clears false-actionable
pk-doctor findings after the v0.26.9 preauth fix and hardens ID token
reservation.

Fixed

• Downgraded historical lexical-token collisions to non-actionable
  info. Existing timestamp/word-pair shorthand collisions now report
  as historical context instead of open migration work, while new ID
  generation continues to avoid reusing reserved lexical tokens. Closes
  #51.
• Reserved lexical tokens across all ID generation paths.
  id-management.generate_id() and the shared processkit ID helper now
  reserve every lexical token already present in the target kind, so
  pair, intent, and generic ID generation all avoid ambiguous reuse.
• Stopped rejected migrations from keeping archive hygiene
  actionable. The pk-doctor applied-migration archive check now counts
  only state = applied migration records, so rejected records can stay
  available as explicit policy dispositions.

Verification

• pk-doctor: 0 errors, 0 warnings, 0 actionable infos
• pk-release-audit --tree=both
• uv run scripts/smoke-test-servers.py
• Focused pytest coverage for ID vocabulary and ID generation

v0.26.9

Patch release for processkit issue #50.

Fixes:

• Adds processkit-runtime-prune to shipped preauth metadata for Claude and Codex MCP permissions.
• Regenerates the MCP manifest with runtime-prune included.
• Updates the manifest generator and release tarball build so source-only shipped MCP configs are included and the packaged manifest is stamped with the target release tag.

Validation:

• uv run scripts/smoke-test-servers.py
• npm --prefix docs-site run build
• uv run src/context/skills/processkit/pk-doctor/scripts/test_doctor.py
• scripts/check-src-context-drift.sh --release-deliverable
• pk-release-audit on src-context
• release tarball checksum verified

v0.26.8

Patch release for processkit issue cleanup.

Fixes

• pk-doctor id_vocabulary now validates configured kind palettes
  instead of warning on the unconfigured default two-word pool. This
  resolves the structurally unresolvable
  id-vocabulary.default-pair-capacity-low warning.
• scripts/processkit-diff.sh now emits normalized affected_files
  rows in JSON and TOML output so migration generators can populate
  Migration frontmatter from the same structured source used for body
  rows.

Validation

• uv run scripts/smoke-test-servers.py
• npm --prefix docs-site run build
• scripts/check-src-context-drift.sh --release-deliverable
• pk-release-audit on src-context
• release tarball checksum verified

v0.26.7 - Semantic ID vocabulary allocation

Patch release for semantic ID vocabulary allocation.

Highlights:

• Adds tagged semantic vocabulary pools and high-volume three-word allocation support in id-management.
• Adds pk-doctor id vocabulary checks for pool capacity, ambiguous lexical tokens, and drift risks.
• Records the accepted vocabulary/RAG allocation decisions and closes the implementation work item.

Validation:

• Release boundary guard passed.
• pk-release-audit passed with 0 errors / 0 warnings.
• Provenance guard passed for v0.26.7.
• Tarball checksum verified.

v0.26.6

Patch release for processkit AGENTS.md hygiene and derived-project startup-policy reconciliation.

Highlights:

• Adds pk-doctor agents_md_hygiene checks for managed AGENTS.md blocks, processkit references, provider pointer thinness, command block schema, and stale guidance patterns.
• Marks AGENTS.md processkit compliance and command sections as pk-managed blocks in the shipped template.
• Updates aibox.toml runtime config to current lazy MCP catalog and tmux PowerKit status symbols.

Validation:

• pk-doctor smoke tests passed.
• pk-doctor JSON contract tests passed.
• smoke-test-servers passed.
• release deliverable guard passed.
• release-audit src-context passed.
• provenance check passed.

v0.26.5 - strict migration policy

v0.26.5 is a patch release that enforces strict schema and storage migration policy.

Fixed:

• pk-doctor no longer treats legacy_known_* schema allowlists as accepted terminal vocabulary.
• Legacy closed-vocabulary drift now requires a data-fix Migration or explicit schema migration.
• TeamMember.default_role is back to canonical ROLE- validation.
• Mixed root/sharded layouts, placeholder IDs, mixed filename policies, and Binding filename drift now report as migration work.
• pk-doctor no longer reports accepted_policy_exception as a terminal machine-readable resolution.
• Event-log, migration-management, pk-doctor, skill-gate, decision-record, and cross-reference guidance now point to declared schema vocabulary and explicit migrations.

Verification:

• pk-doctor tests passed in context/ and src/context/.
• smoke-test-servers passed.
• src-context release audit passed with 0 errors and 0 warnings.
• release tarball checksum verified.

Closes #47.

v0.26.4 - schema compatibility regression fix

v0.26.4 is a patch release that restores schema compatibility for existing aibox dogfood contexts after the v0.26.3 integration regression.

Fixed:

• pk-doctor schema_vocabulary now accepts declared legacy_known_types, legacy_known_kinds, and legacy_known_event_types while keeping unknown-value checks intact.
• Shipped schemas now grandfather legacy feature WorkItems and historical dogfood LogEntry event types including release.shipped, schema.migrated, and migration.verified.
• TeamMember.default_role now accepts migrated legacy ROLE-* IDs with datetime/word-pair shapes.

Verification:

• pk-doctor tests passed in context/ and src/context/.
• smoke-test-servers passed.
• src-context release audit passed with 0 errors and 0 warnings.
• release tarball checksum verified.

Closes #46.

v0.26.3 - runtime prune MCP

v0.26.3 is a patch release that adds the runtime-prune processkit skill and MCP server for provider-neutral aibox-managed cleanup workflows.

Added:

• Added analyze_disk_usage, plan_prune, and apply_prune MCP tools.
• apply_prune requires an exact confirmation token and delegates destructive work to aibox prune.

Closes #45.

v0.26.2 - doctor action queues and migration listing fixes

v0.26.2 is a patch release that fixes the remaining GitHub issue
queue for migration listing, doctor actionability, and session-start
maintenance prompts.

Added

• Added pk-doctor actionability metadata. Each doctor finding now
  includes action_required, action_kind, default_agent_action,
  requires_user_confirmation, and acceptable_resolution, with
  aggregate action_totals surfaced alongside severity totals. This
  prevents actionable INFO findings from disappearing behind a clean
  0 WARN gate. Closes #43.
• Added pk-resume action-queue guidance. Status briefings now treat
  severity and actionability as separate queues, require dispositions
  for actionable doctor findings, and require explicit apply/reject/
  inspect/defer decisions for pending or in-progress migrations. Closes
  #44.
• Imported the expanded skill catalog from the current aibox sync.
  The dogfood tree now includes the generated data-ai, design, devops,
  documents, engineering, and product skill families plus corresponding
  command adapters.

Fixed

• Fixed lazy gateway serialization for empty structured list
  results. Catalog-backed gateway tools now preserve FastMCP
  structured output and return a text fallback when the unstructured
  content list would otherwise be empty. This keeps
  list_migrations(state="pending") valid when no Migration entities
  exist or when non-entity Markdown briefings sit in pending/. Closes
  #42.
• Resolved and archived generated migration records. The invalid
  processkit/runtime migrations from the aibox sync were rejected with
  integrity reasons, the runtime drift migration was applied by
  preserving local edits, and all three terminal migration entities were
  archived via context-archiving.

Verification

• python3 -m py_compile context/skills/_lib/processkit/gateway/lazy.py context/skills/_lib/processkit/gateway/registry.py src/context/skills/_lib/processkit/gateway/lazy.py src/context/skills/_lib/processkit/gateway/registry.py
• uv run --with mcp --with pyyaml --with jsonschema --with httpx --with sqlite-vec --with pytest pytest context/skills/processkit/processkit-gateway/scripts/test_gateway.py context/skills/processkit/migration-management/scripts/test_migration_management.py -q
• uv run --with mcp --with pyyaml --with jsonschema --with httpx --with sqlite-vec --with pytest pytest src/context/skills/processkit/processkit-gateway/scripts/test_gateway.py src/context/skills/processkit/migration-management/scripts/test_migration_management.py -q
• python3 -m py_compile context/skills/processkit/pk-doctor/scripts/doctor.py context/skills/processkit/pk-doctor/scripts/checks/common.py context/skills/processkit/pk-doctor/mcp/server.py src/context/skills/processkit/pk-doctor/scripts/doctor.py src/context/skills/processkit/pk-doctor/scripts/checks/common.py src/context/skills/processkit/pk-doctor/mcp/server.py
• uv run context/skills/processkit/pk-doctor/scripts/test_pk_doctor_json.py
• uv run context/skills/processkit/pk-doctor/scripts/test_doctor.py
• uv run src/context/skills/processkit/pk-doctor/scripts/test_pk_doctor_json.py
• uv run src/context/skills/processkit/pk-doctor/scripts/test_doctor.py
• uv run scripts/smoke-test-servers.py

---

v0.26.1

v0.26.1 is a patch release focused on keeping status/reporting
workflows small, making release health checks directly available through
the gateway, and closing the local storage hygiene issues tracked in
GitHub #39-#41.

Added

• Added bounded result windows for database-backed MCP queries.
  Entity, search, event, error, TeamMember, RoleSlot, and ID-listing
  queries now clamp oversized limits; unfiltered event queries return a
  smaller default window and truncate long summaries to avoid noisy,
  token-heavy responses.
• Added active-only default migration listing. list_migrations()
  now returns pending and in-progress migrations by default; historical
  applied/rejected migrations are available through explicit state
  filters.
• Added entity_storage_hygiene to pk-doctor. The new check audits
  local context storage policy, unmanaged host artifacts, legacy roots,
  lifecycle buckets, filename policy drift, placeholder timestamps, and
  documented TeamMember privacy/local-state exceptions.
• Added the pk-doctor MCP config and gateway catalog exposure.
  run_pk_doctor is now surfaced as a direct read-only MCP tool, and
  run_pk_release_audit, list_entities, and get_entity_by_path
  are present in the generated gateway catalog.
• Expanded positive two-word ID naming pools. The default adjective
  and noun pools now provide 14,400 base combinations per kind before a
  collision suffix is needed.

Changed

• pk-resume, standups, and status updates now check GitHub state.
  In GitHub-backed repositories they include open issues and PRs when
  gh is available, and pk-resume includes pk-doctor totals as part
  of the session-start health check.
• Team command adapters were refreshed from the current source
  skills. .agents/skills and .claude/skills now match the updated
  team-creator command bodies.
• Local release workspace hygiene was tightened. Old diagnostics and
  root-level .gitignore.example scaffolding were removed from the
  repository, and applied migration history was archived through the
  context archive workflow.

Fixed

• Closed GitHub issues #39, #40, and #41. pk-doctor now separates
  local entity storage hygiene from template freshness and reports
  actionable storage drift without flagging documented flat or
  grandfathered layouts.
• Resolved pending local migrations. The hidden root-level
  v0.25.8 -> v0.26.0 Migration was applied through MCP and then
  archived with the rest of the applied migration history.

Verification

• uv run context/skills/processkit/pk-doctor/scripts/test_doctor.py
• uv run context/skills/processkit/pk-doctor/scripts/doctor.py
• gh issue list --state open --limit 20 --json number,title,url

---