project-tiny-context-harness 0.11.0 → 0.12.0

package/migrations/README.md

@@ -1,191 +1,32 @@
-# Migrations
+# Schema-5 retirement (0.12.0)
 
-Schema migrations for Harness config and managed file layout belong here.
+Update mode: **upgrade-required**. Automatic migration is limited to schema 4 at the package 0.11.0 baseline, commit `8cf2391295cbec271f148dc44f074bedeea767b5`. It retires workflow software; it does not certify that every project rule or historical resource has been converted.
 
-The completed 0.10.x line adds opt-in, dry-run-first `context register`, `context move` and
-`context transaction status|rollback|complete`. Current writes use
-`context-mutation-journal-v3`, whose immutable generations bind each NFC logical
-key to its exact physical path while retaining complete mode/file-identity
-endpoint snapshots and the shared Active Authority lock.
-`context-mutation-journal-v2` remains recovery-readable with equal logical and
-physical paths. An unfinished pre-v2 journal is deliberately not guessed or
-rewritten: use the matching older package to recover it, inspect/restore the
-recorded bytes manually, or obtain project-owner recovery direction before a new
-mutation. They do not migrate existing
-projects or reinterpret legacy policies. Register losslessly appends one
-recoverable Context node. Move preserves one registered Context/Area owner,
-structured default-child references and explicit local Context-Markdown links;
-unstructured exact references or an incomplete bounded scan block apply. Both
-use current-byte/mode/file-identity endpoint comparison, candidate-tree
-validation and one logical recoverable journal published as immutable
-digest-linked generations. Recovery and Long-Task Authority changes serialize
-through the existing Active Authority lock. There is no automatic policy
-conversion, force bypass, hidden TOML normalization, hostile-writer mutex or
-claim of cross-file physical atomicity.
+1. Inspect with `ty-context upgrade --check`. Stop relevant old host sessions before `ty-context upgrade --sessions-stopped`. A file lock cannot prove a host session has stopped.
+2. Settle any old Context transaction before changing schema/manifest semantics or removing the old engine. Use the installed compatible 0.11.0 tool, never `@latest`. Complete or roll back; if an old binding conflicts, legitimately end/abandon it first. No Final Gate or lock deletion is required. Unreadable journals block with a diagnostic.
+3. Reconcile modified old Skills, worker profiles or Hooks, custom managed entries and active build dependencies. Exact old file hashes identify removable package assets. Mixed Hook entries and exact Makefile includes are edited locally, keeping user portions. Other worktree records and configuration keys are outside the migration scope.
+4. Migration saves original bytes and planned replacements under `tmp/ty-context/upgrade-backups/<id>.json`, then writes `tmp/ty-context/upgrade-schema-5.json`. It publishes assets/manifest/configuration before the startup entry, checks current files and defaults, and retires only the target worktree binding. Normal writes refuse pending publication. These files are outside automatic Skill/default Context loading.
+5. An interrupted migration resumes through the same explicit command. Do not remove a pending journal, overwrite a conflict, or edit its planned file contents. Original and pending plans must agree. Preserve the backup while resolving reported file conflicts. Filesystem replacement is not atomic across the whole installation.
+6. Start a new host session after successful migration. Already-loaded instructions do not disappear when their file is removed.
 
-Version 0.10.1 adds `ty-context context create --path <path> --role <role>`.
-It writes one TODO-only, unregistered Markdown scaffold and does not modify
-`project_context/context.toml` or the default footprint. Existing files,
-registered paths, unsafe/link-traversing paths and concurrently appearing
-destinations are refused rather than overwritten. Replace every TODO with
-durable owner facts before a separate registration step; the unchanged
-scaffold intentionally does not satisfy registered-Context recoverability.
-This is a new opt-in command and requires no consumer migration.
+## Default-body compatibility
 
-Version 0.11.0 adds an opt-in sparse Design Authority bundle. Existing projects
-with only an unmarked root `DESIGN.md` remain valid one-file closures and are
-not rewritten. Adoption inserts the exact
-`<!-- ty-context-design-authority-format: bundle-v1 -->` declaration as the
-first non-empty Markdown body line after supported YAML front matter and creates
-`design_system/authority.manifest.json` in the same explicit DSA/Authority
-Revision. Generated `design_system/tokens.json` and subordinate owners remain
-explicit adoption outputs, not upgrade migration. Marker and manifest are a
-strict pair: either half missing, renamed, empty or invalid fails closure. DRA
-recovery/handoff and Long-Task bindings use the complete entry plus closure
-digest; a raw-file-only legacy binding fails closed and must be regenerated or
-rebound. A deliberate return to one-file Authority removes both halves only
-through explicit DSA adoption/Authority Revision. Static inspection can enforce
-the current pair and stale bindings, but a fresh checkout containing neither
-half has no authenticated historical adoption registry. Upgrade never infers
-Token values, owner splits, candidate selection, adoption or a human revision.
+The exact normalized body path set is compared under old and new selection semantics. The migration adds the formerly implicit architecture file to nonrecursive `default_files`; it does not promote that node into a `default_children` traversal root. Explicit default-node traversal, legacy policies and referenced `never-default` children keep their prior behavior. Manifest routing metadata is separate from default body text. Any addition/removal is a blocking concrete path diff. Later user-authorized Context pruning is a separate change.
 
-Version 0.10.0 includes `context-units-to-context`. Explicit `upgrade` applies
-only when every retired `[[context_units]]` table is a provably simple
-single-line TOML table. It renames the table and removes only retired `id` and
-`area` fields while preserving all untouched bytes and the existing EOL.
-Complex/mixed-EOL input, unknown legacy fields and current/legacy path
-conflicts are `manual_required`; ordinary `sync` never runs this migration.
+## Retained resources and review scope
 
-Version 0.6.0 includes `long-task-v1-retirement`. It safely removes the
-retired repo-local Hook, reports a legacy active projection as
-`manual_required`, and deliberately does not import V1 progress or receipts
-into the V2 Claim/Evidence authority.
+Historical Contracts, Receipts and ordinary project files remain unchanged and are not executed or universally converted. Ordinary Markdown and direct JSON/YAML fields remain readable with their explicit local paths. No arbitrary structured-ID/cross-file adapter is provided. The migration does not expand compact forms, evaluate symbolic resources or restore old compiler/observer semantics.
 
-Version 0.7.2 includes `long-task-v2-semantic-drift-authority`. It reports a
-conventional `.long-task/delivery-contract.yaml` as `manual_required` when the
-V2 file predates explicit Stage, target-profile/root-runtime, journey-scenario,
-success/degradation, evidence-capability or external-impact authority. Those
-meanings must be re-authored from Source; migration never infers them from old
-Progress or Receipts.
+The bounded review reads current Context/DESIGN and root AGENTS overrides, follows explicitly declared `ty-context-controlling-source` local dependencies, and reports ordinary local Markdown links for review. It inspects root and manifest-Area `package.json` scripts and retained Makefile recipes for literal retired CLI calls. Named symbolic handoff/IR inputs required by controlling declarations block until necessary unique requirements, adopted decisions and locators are extracted or replaced with provenance. Ordinary historical links are advisory, not existence blockers. Inspection of an individual resource over 2 MB is explicitly unresolved rather than silently partial.
 
-## Version 0.9.0 complete-delivery authority
+Current task references, selected resources, compact IDs, arbitrary CI/shell indirection and other build systems require specific review by the upgrading agent/user. For a known dependency, report the file, reference and unavailable content; do not claim information migration complete until it is addressed. Do not regenerate design or infer requirements to replace original decisions. No network checks, resource generation, historical-proof repair or second compiler is introduced.
 
-Version 0.9.0 retains the `long-task-delivery-v2` Contract marker. A Contract
-that omits `task.target_profile.completion_authority` is parsed as
-`machine_only`, preserving the older fail-safe machine-only meaning, and
-Preflight emits the actionable
-`completion_authority_legacy_default_machine_only` warning. New authoring
-should state either `machine_only` or `declared_authorities` explicitly.
+## Journal and old-binary compatibility
 
-Material Source compilation now derives fragment and anchor projections below
-the Source-item level. Every `supporting_basis` fragment/anchor projection must
-name at least one known delivery-semantic Fact; an empty or integrity-only Fact
-set is rejected. Legacy single-fragment Source-item projections that already
-carry delivery Fact refs remain readable through deterministic projection, but
-new authoring should classify every multi-fragment row explicitly. Superseding
-Source must carry the same-domain replacement meaning in a delivery-semantic
-Fact; migration does not invent that Fact or infer precedence.
+Current Context mutations use `context-mutation-journal-v3`; the retained recovery parser also understands `context-mutation-journal-v2`. Unrecognized pre-v2 journals require manual recovery using a matching version. This format compatibility does not authorize a schema-5 writer to recover an unfinished schema-4 installation: settle that transaction with its compatible old tool first.
 
-Changing to `declared_authorities` is a semantic owner decision, not a
-mechanical migration. Re-author every blocking external route from Source with
-its declared actor/owner, target, environment, keyed Given/When scenario,
-evidence requirements and exact per-obligation Claim/applicability/Fact/proof/
-method/capability/Expected-authority decomposition. Migration must not infer an
-actor, authenticate a real-world identity, weaken Expected, collapse rows into
-an aggregate Boolean, or silently convert unsupported machine proof to an
-external judgment.
+New sync/init/mutation/export writes reject old, malformed or unfinished schema configuration. Old command tombstones return failure. New binaries cannot retroactively fix old binaries: baseline old `enable` may write a profile before reaching its schema guard. Schema-5 configurations containing retired profiles/modularity are rejected rather than interpreted as runnable old capabilities. Baseline bare init can reselect .codex before checking the formerly selected .agent schema, successfully reinstalling old startup assets. Fixed-root init and sync reject schema 5, but that does not protect root reselection. New writes reject the resulting schema-4 installation and do not treat it as current. This limitation is tested with the actual old binary; no daemon or retrospective guarantee is claimed. Do not mix package versions after retirement.
 
-New Final Gates emit `long-task-final-receipt-v3` with one of
-`machine_accepted`, `delivery_accepted`, `blocked_external` or `needs_work`.
-Version 2 Receipts remain readable for audit only. They are not current
-acceptance evidence, cannot be upgraded by appending an External Confirmation
-record and must be followed by a current Compile and complete Final Gate.
-Legacy `machine_accepted_external_pending` remains compatibility-read-only and
-cannot clear Active Authority or close. Only fresh `machine_accepted` and
-`delivery_accepted` may close.
+## Verification boundary
 
-External fulfillment uses the strict
-`long-task-external-confirmation-record-v1` input and the `external prepare`,
-`submit`, `status` and `revoke` commands. Record hashes prove local integrity,
-not signatures or actor authentication. Relevant-input changes invalidate only
-when independence is soundly derivable; otherwise the whole record becomes
-stale. These rules make the release `manual-required` for affected active
-Long-Task consumers. No generic Browser, Native or Device observer SDK is part
-of this migration.
-
-## Admitted-observation/runtime-TCB revision
-
-The next Long-Task compatibility boundary intentionally does not provide a
-mechanical Contract rewrite. Machine evidence that depends on a custom Oracle,
-a project-submitted runtime/interaction/state/verdict row, a runner-created or
-runner-modified carrier, an indirect wrapper, an unsupported target family or
-a Counterfactual with no package-admitted actual must be re-authored explicitly.
-Migration must never silently:
-
-- replace a custom Oracle with a package Oracle;
-- turn machine proof into `external_confirmation`;
-- replace a verifier wrapper with a product root entrypoint;
-- infer a static structure carrier from `Binding`/`input_paths` alone; or
-- infer a process runtime dependency from `input_paths`, runner or verification membership;
-- invent a Source-backed execution target or production Binding for an old root; or
-- preserve a known false-acceptance path as compatibility behavior.
-
-Greenfield authoring may declare an exact process root, argv repository file
-or production carrier as `planned`. Its exact Source/Binding declaration is
-stable through Preflight/Compile and later materialization, but Final Gate
-requires every compiled closure member to exist. A glob, a manifest sibling or
-an undeclared repository-file argv token cannot substitute; migration does not
-invent these exact declarations or parse language dependencies.
-
-Preflight and doctor should report the narrow failing obligation and an
-actionable manual route through diagnostics such as
-`machine_observer_not_admitted`,
-`semantic_fact_machine_observer_not_admitted`,
-`unsupported_observer_requires_external_confirmation`,
-`custom_oracle_machine_completion_forbidden`,
-`static_observation_not_in_pre_run_snapshot`,
-`static_observation_changed_by_runner`,
-`process_observer_direct_root_required`,
-`process_observer_root_invocation_required`,
-`process_observer_root_argv_mismatch`,
-`process_observation_input_changed_by_runner`,
-`process_runtime_input_expected_authority_forbidden`,
-`process_runtime_input_verification_role_forbidden`,
-`process_runtime_input_evidence_role_forbidden`,
-`process_root_source_binding_required`,
-`process_root_production_binding_required`,
-`process_root_source_identity_mismatch`,
-`process_runtime_carrier_exact_path_required`,
-`process_runtime_input_missing`,
-`process_runtime_closure_identity_mismatch`,
-`legacy_target_runtime_non_authoritative`,
-`counterfactual_admitted_observation_required`,
-`counterfactual_runtime_reachability_unproven` and
-`project_submitted_verdict_disagrees_with_harness`. The user or Agent must then
-choose an admitted static/direct-process observation or the existing blocking
-External Confirmation. That semantic choice is `manual_required`, not ordinary
-`sync` or a migration-registry inference; release guidance must therefore use
-the `manual-required` update mode for affected consumers.
-
-When the whole affected Outcome is unsupported, re-author it explicitly as an
-external-only route: set `success_path_required: false`, remove machine Checks,
-bind every ordinary/global and Semantic Fact Claim through exact
-`impact_claims`, keep every Semantic Fact proof's explicit `confirmation_ref`,
-and ensure a `blocks_target: true` confirmation impacts each
-Stage Gate result Claim. A non-blocking confirmation or one missing gate-result
-lineage cannot replace a Stage Gate, and migration never invents these choices.
-The valid route ends as `blocked_external`, not either machine-accepted status.
-
-Changing verifier identity, Compile admission policy or the observation TCB
-invalidates the prior Active Authority. The Contract must be recompiled or
-adopted through an explicit Authority Revision; old Progress, evidence and
-Receipt are not reusable on the new boundary. Public
-`long-task-check-result-v3` remains the compatibility payload, but its
-project-authored actual/pass/verdict/runtime fields are non-authoritative. The
-Harness-owned process adapter captures the separate
-`ty-context-product-observation-v1` envelope from bounded stdout inside an
-isolated frozen runtime-closure snapshot. It exposes no output-path, challenge
-or protocol environment variable to the child; any execution nonce remains
-internal host-attestation data. No v4 is introduced by this revision; a future
-v4 requires a distinct need to carry a live browser/native session handle
-across processes.
+Tests exercise real baseline init, interrupted move plus compatible complete/rollback, exact default sets, interrupted upgrade resume, tampered plans, mixed user configuration, modified/additional Skill files, current resource/build dependencies and worktree-key isolation. Product behavior and nonmanaged project-rule correctness remain outside structural validation. Platform and tarball results must be reported from actual runs, not inferred from these notes.

package/migrations/schema-4-owned-assets.json

@@ -0,0 +1,170 @@
+{
+  "baseline": "8cf2391295cbec271f148dc44f074bedeea767b5",
+  "package_version": "0.11.0",
+  "files": [
+    {
+      "path": ".codex/agents/long-task-implementation.toml",
+      "sha256": "dd6f0f29ca9f452e1fcf934f9ec472a6ca4fd66a6f3be53c42823fdcb5f2aa7d"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/architecture.md",
+      "sha256": "05727cd0b4f2308811be59936798ac9322cc8f1009e95e99705c679d72b1ba0c"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/area.md",
+      "sha256": "64d2bf9eaaf8bed1bda4db5567172afca19298cd6579243a423d56b31214335b"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/context.toml",
+      "sha256": "1a1c49d3613fd8e7b993460cca56a2f3d9239c536ef726e88c8051b177706929"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/deployment.md",
+      "sha256": "d9977aae6b344b9dbee1df36909af6013f4569652c303758ce3ccb559a29faa4"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/global.md",
+      "sha256": "af1356447e9697408fe0d42c0a060cd9f8d89d00ffde3c29b11bdef97247e193"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/product-surface-contract.md",
+      "sha256": "475f7f82c383f8a0bc76480d95fb72774107b42f6f478b589bb1fc8863a32bca"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/screen-contract.md",
+      "sha256": "9470eabe4c7ee9982c2eb4d8cc21d2b9ecc3cc286249435c1e0a24bb0a85be06"
+    },
+    {
+      "path": "{root}/ty-context-managed/context_templates/verification.md",
+      "sha256": "290e1f7d0b3624906771d12a8b79933fd2faf143a51749787c063aa3e5893208"
+    },
+    {
+      "path": ".github/workflows/harness.yml",
+      "sha256": "c640ae4ab1eb4573277771f2560c5bac1637deba50ce6a44ce1888afe96139ae"
+    },
+    {
+      "path": "{root}/ty-context-managed/make/ty-context.mk",
+      "sha256": "97f3051e953ef06ab379c75f77931dea9887e55ae0eb5e9e6938de7b43cab14b"
+    },
+    {
+      "path": "{root}/skills/context_development_engineer/references/engineering-design-reasoning.md",
+      "sha256": "ff86970f6a43458c0f1373cbc42f2bbe97d5d41e49b82a865cf30f1a0c90789e"
+    },
+    {
+      "path": "{root}/skills/context_development_engineer/SKILL.md",
+      "sha256": "a4c52f9244f6c5df22c968e9712abe35b3c121013b9e9f647ee8aa91c327e9d0"
+    },
+    {
+      "path": "{root}/skills/context_full_project_export/SKILL.md",
+      "sha256": "dc4af5730e54bdac26c64999b3370f7cb688bb33270f3f1580c0015200ac0f5d"
+    },
+    {
+      "path": "{root}/skills/context_harness_upgrade/SKILL.md",
+      "sha256": "67ab784c021bb2a015781e5567f5c010531b6dfd4d58f7fa162a36650b9d5fc5"
+    },
+    {
+      "path": "{root}/skills/context_product_plan/SKILL.md",
+      "sha256": "f31e6fbfd5c1d7f1a6b83d0dfdab8c80d364ead3d028fd190dcc2d29d5ccdf75"
+    },
+    {
+      "path": "{root}/skills/context_surface_contract/SKILL.md",
+      "sha256": "b159b7c4674fec3194e7835bc0393dc4eb8946fc5721253798482f69bd9a9e87"
+    },
+    {
+      "path": "{root}/skills/context_uiux_design/references/task-uiux-analysis.md",
+      "sha256": "4e5806a1c2d47163a8f9bf964f0c9e0ddb5793efe3732e79cfb69d85037bf1b4"
+    },
+    {
+      "path": "{root}/skills/context_uiux_design/SKILL.md",
+      "sha256": "dabf9f98c7b0990f2e42d45549a15c7852597fd14b56e7fd6122a0d7ff230d5f"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/authority-delta-assessment.md",
+      "sha256": "d68b9d478ac6e0563384073a76f39624705138e7e036daf3429e180408e848d5"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/downstream-handoff.md",
+      "sha256": "019c691b4fe01adf7f8a81b784c51f209fc0cf975fd491ee876dfd98a4fa78e2"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/formal-selected-web-app-handoff.md",
+      "sha256": "b436e4b555e206248d5d93072c85bb2336adb80a9e75b4af30d41793c5252005"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/implementation-feasibility.md",
+      "sha256": "6a9af7a54c42bd173a483a1c45a6681338cd9af521f76c8ea4f8b0cfa8b41dce"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/open-design-provider.md",
+      "sha256": "cb5feb506d71c551ae07bd7340b95bdb1e52224ab9161ce9da1793dbe6dd7613"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/recovery-and-writeback.md",
+      "sha256": "524cd13e0d5b94e784b596b7f2985479dac8760d000c72424ef0028ead32578c"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/references/resource-selection.md",
+      "sha256": "b1690e2992d4a8eaf92d3ede8de94161cc96fdd6a252545aaae13ec0bf8762c8"
+    },
+    {
+      "path": "{root}/skills/design-resource-authoring/SKILL.md",
+      "sha256": "555523161ee6095f8a387e31a9b54af5205490537397a256a67d6366ffa770b2"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/agents/openai.yaml",
+      "sha256": "a562fd7aaf1e7e329bb2dcf7ce4ba25aa815dc858cd63f069b61aa6307bafea5"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/references/authority-adoption.md",
+      "sha256": "4bdd35e2775ddc0289de93a811febb53b25716e55dc9e44c590725afb22d170f"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/references/open-design-design-system-provider.md",
+      "sha256": "621d633ab76b375a08dd02e07e39e16de18c67bfb81839a184a379e0a4968845"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/references/showcase-projection.md",
+      "sha256": "93d58db1088e9926397c8c6327509dec3ecced212a52164f1597aa9bca2e2cd8"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/references/system-design-reasoning.md",
+      "sha256": "22574a6c773c0ad53194db43439ced2b42e16c3906bcd9cfd2a4174d5494b5bf"
+    },
+    {
+      "path": "{root}/skills/design-system-authoring/SKILL.md",
+      "sha256": "0e2473636ae12646da3a28f4c44fbd0efabd22de9c2a5564779624c257dbb088"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/agents/openai.yaml",
+      "sha256": "f3458a75b27013d2ce3a0875a667cd89304313047375ece3ba90e36ac5440d06"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/references/authority-lifecycle.md",
+      "sha256": "065e36fb60b326ad81e512e92ce0215ac52332021c8df29fbcb4241e23666350"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/references/contract-authoring.md",
+      "sha256": "ee490c8335b4eae04359470bb12f3f65cfa6dd21cbab65dc5ce067d7029cf84e"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/references/evidence-design.md",
+      "sha256": "02b517ecdad4edda75154f1e1b2beac39a0a5b5767a564ab473bf466b638adc6"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/references/source-authoring.md",
+      "sha256": "d85ae52d44ba48ffcaf1026814755e41b0f5a4a348cfe2f9f819e847fad450c0"
+    },
+    {
+      "path": "{root}/skills/long-task-workflow/SKILL.md",
+      "sha256": "92055aa78371f4649b1a59b7e7c069ea845cdcce23da580e3db4cb86cfb1320a"
+    },
+    {
+      "path": "tools/context_rules.json",
+      "sha256": "1c373fcb5510f6c38a772b638fe01a8b8e96b2230540a1514811819e90b01086"
+    },
+    {
+      "path": "tools/validate_context.py",
+      "sha256": "a263a6a911a7a51e7387184061d6a166e620313f12e7e55d3813b8322aad96ad"
+    }
+  ]
+}

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "project-tiny-context-harness",
-  "version": "0.11.0",
-  "description": "Minimal project memory and validation harness for AI coding agents.",
+  "version": "0.12.0",
+  "description": "Durable project facts and a short development contract for coding agents.",
   "license": "MIT",
   "author": "Seven128",
   "homepage": "https://github.com/Seven128/project-tiny-context-harness#readme",
@@ -48,33 +48,18 @@
     "source-mappings.yaml"
   ],
   "scripts": {
-    "build": "node ../../tools/package_build_fingerprint.mjs build",
+    "build": "node ../../tools/build_package.mjs",
     "typecheck": "tsc -p tsconfig.json --noEmit",
-    "test:default:built": "node ../../tests/ty-context/run-package-suite.mjs default",
-    "test:default": "npm run build && npm run test:default:built",
-    "test:built": "npm run test:default:built && npm run test:long-task-workflow:built",
-    "test:trust:built": "npm run test:default:built && npm run test:long-task-trust:built",
-    "test:trust": "npm run build && npm run test:trust:built",
     "pretest": "npm run build",
-    "test": "npm run test:built",
-    "test:workflow-default:built": "node --test --test-concurrency=1 ../../tests/ty-context/workflow-contract-routing.test.mjs",
-    "test:delivery-contract:built": "node --test --test-concurrency=1 ../../tests/ty-context/long-task-compact-authoring.test.mjs ../../tests/ty-context/long-task-compact-semantic-carrier.test.mjs ../../tests/ty-context/long-task-authoring-claims.test.mjs ../../tests/ty-context/long-task-authoring-preflight.test.mjs ../../tests/ty-context/long-task-authority-authoring-fields.test.mjs ../../tests/ty-context/long-task-playwright-ac-evidence.test.mjs ../../tests/ty-context/long-task-delivery-parser.test.mjs ../../tests/ty-context/long-task-delivery-compiler.test.mjs ../../tests/ty-context/long-task-delivery-risk.test.mjs ../../tests/ty-context/long-task-claim-coverage.test.mjs ../../tests/ty-context/long-task-closure-invariants.test.mjs ../../tests/ty-context/long-task-source-authority-closure.test.mjs ../../tests/ty-context/long-task-evidence-sensitivity-policy.test.mjs ../../tests/ty-context/long-task-global-evidence-sensitivity.test.mjs ../../tests/ty-context/long-task-source-risk-binding.test.mjs ../../tests/ty-context/long-task-non-completing-source.test.mjs ../../tests/ty-context/long-task-playwright-trust-boundary.test.mjs ../../tests/ty-context/long-task-planned-counterfactual.test.mjs ../../tests/ty-context/long-task-public-contract-example.test.mjs ../../tests/ty-context/long-task-release-tarball-contract.test.mjs ../../tests/ty-context/long-task-semantic-assurance-closure.test.mjs ../../tests/ty-context/long-task-semantic-drift-closure.test.mjs ../../tests/ty-context/long-task-semantic-drift-lifecycle.test.mjs ../../tests/ty-context/long-task-workspace-scope.test.mjs",
-    "test:delivery-contract": "npm run build && npm run test:delivery-contract:built",
-    "test:long-task-workflow:built": "node ../../tests/ty-context/run-package-suite.mjs long-task",
-    "test:long-task-workflow": "npm run build && npm run test:long-task-workflow:built",
-    "test:long-task-trust:built": "node ../../tests/ty-context/run-package-suite.mjs long-task-trust",
-    "test:long-task-trust": "npm run build && npm run test:long-task-trust:built",
-    "test:long-task-performance": "npm run build && node ../../tests/ty-context/long-task-performance.mjs",
+    "test": "node ../../tests/ty-context/run-package-suite.mjs",
+    "test:built": "node ../../tests/ty-context/run-package-suite.mjs",
     "prepack": "npm run build"
   },
   "engines": {
     "node": ">=24"
   },
   "dependencies": {
-    "@google/design.md": "0.4.0",
-    "impeccable": "^3.6.0",
     "mdast-util-from-markdown": "^2.0.3",
-    "re2js": "2.8.6",
     "smol-toml": "1.8.0",
     "toml-eslint-parser": "1.0.3",
     "yaml": "^2.9.0"
@@ -82,5 +67,14 @@
   "devDependencies": {
     "@types/node": "^24",
     "typescript": "^7.0.2"
+  },
+  "main": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js"
+    },
+    "./package.json": "./package.json"
   }
 }

package/source-mappings.yaml

@@ -2,45 +2,9 @@ source_mappings:
   - source: ".codex/ty-context-managed/agents/AGENTS_CORE.md"
     target: "packages/ty-context/assets/agents/AGENTS_CORE.md"
     mode: "copy-file"
-  - source: ".codex/ty-context-managed/agents/long-task-implementation.toml"
-    target: "packages/ty-context/assets/agents/long-task-implementation.toml"
-    mode: "copy-file"
-  - source: "README.md"
-    target: "packages/ty-context/assets/README.md"
-    mode: "copy-file"
-  - source: "README.zh-CN.md"
-    target: "packages/ty-context/assets/README.zh-CN.md"
-    mode: "copy-file"
-  - source: ".codex/ty-context-managed/context_templates"
-    target: "packages/ty-context/assets/context_templates"
-    mode: "copy-tree"
-  - source: ".codex/ty-context-managed/skills"
-    target: "packages/ty-context/assets/skills"
-    mode: "copy-tree"
-  - source: ".codex/ty-context-managed/make/ty-context.mk"
-    target: "packages/ty-context/assets/make/ty-context.mk"
-    mode: "copy-file"
-  - source: ".codex/ty-context-managed/minimal_tools"
-    target: "packages/ty-context/assets/tools"
-    mode: "copy-tree"
-  - source: ".codex/ty-context-managed/minimal_tools/validate_context.py"
-    target: "tools/validate_context.py"
-    mode: "copy-file"
   - source: ".codex/ty-context-managed/minimal_tools/context_rules.json"
-    target: "tools/context_rules.json"
+    target: "packages/ty-context/assets/tools/context_rules.json"
     mode: "copy-file"
-  - source: ".github/workflows/harness.yml"
-    target: "packages/ty-context/assets/github/harness.yml"
-    mode: "copy-file"
-  - source: "tools/formal_process_supervisor_native_types.cs"
-    target: "packages/ty-context/assets/runtime/windows-job-supervisor/formal_process_supervisor_native_types.cs"
-    mode: "copy-file"
-  - source: "tools/formal_process_supervisor_native_run.cs"
-    target: "packages/ty-context/assets/runtime/windows-job-supervisor/formal_process_supervisor_native_run.cs"
-    mode: "copy-file"
-  - source: "tools/formal_process_supervisor_native_helpers.cs"
-    target: "packages/ty-context/assets/runtime/windows-job-supervisor/formal_process_supervisor_native_helpers.cs"
-    mode: "copy-file"
-  - source: "tools/windows_job_process_supervisor.ps1"
-    target: "packages/ty-context/assets/runtime/windows-job-supervisor/windows_job_process_supervisor.ps1"
+  - source: "README.md"
+    target: "packages/ty-context/README.md"
     mode: "copy-file"