@litfamily/litopencode 1.0.0

package/.gitattributes

@@ -0,0 +1,8 @@
+skills/frontend-ui-ux/references/design/** -text
+skills/frontend-ui-ux/references/designpowers/** -text
+skills/frontend-ui-ux/references/perfection/** -text
+skills/frontend-ui-ux/references/ui-ux-db/** -text
+skills/frontend-ui-ux/references/_canonical-corpus/LICENSE -text
+skills/frontend-ui-ux/references/_canonical-corpus/ATTRIBUTION.md -text
+skills/frontend-ui-ux/references/_canonical-corpus/LICENSE-Apache-2.0.txt -text
+docs/assets/readme/JetBrainsMono-OFL.txt whitespace=-blank-at-eol

package/ATTRIBUTION.md

@@ -0,0 +1,56 @@
+# ATTRIBUTION / NOTICE
+
+LitOpenCode is authored by this project. One design idea is adapted from a third-party project and
+is credited below. No third-party source is vendored here — only the verification idea is adapted
+into a prompt contract.
+
+---
+
+## 1. insane-research (fivetaku) — inspiration for the claim-graph verification gate
+
+The claim-graph verification gate in `skills/litresearch/SKILL.md` (Phase 3b: a data-flow lock where
+synthesis may assert a high-risk non-code claim only after it clears **≥ 2 independent source
+domains + 1 counter-search + 1 primary source**, and is otherwise abstained into an
+unresolved/refuted annex) is inspired by the data-flow-lock verification design in
+**insane-research** by fivetaku.
+
+- Source: https://github.com/fivetaku/insane-research
+- License: MIT (declared in the project's `README.md`).
+- **What is adapted:** the IDEA only — a verification gate whose output is the sole allowlist the
+  synthesis draws from, so skipping verification leaves nothing to synthesize. No insane-research
+  code is copied or redistributed. The upstream gate is a Python checker (`validate_ledger.py`);
+  LitOpenCode expresses the same concept as a runtime-agnostic prompt and message-text contract,
+  because the OpenCode host guarantees no Python.
+
+```
+MIT License
+
+Copyright (c) 2026 fivetaku
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+```
+
+---
+
+## Maintenance rule
+
+The claim-graph gate and this notice move together. If Phase 3b is ever removed from
+`skills/litresearch/SKILL.md`, this section may be removed with it; while the gate ships, this
+notice must ship too. Porting the gate without the notice is a licence violation, which is why the
+two landed in a single change.

package/CHANGELOG.md

@@ -0,0 +1,267 @@
+# Changelog
+
+Notable user-visible changes are recorded here. Registry and Git release operations
+are tracked separately from this product history.
+
+## Unreleased
+
+## 1.0.0
+
+- Prepare `@litfamily/litopencode` for independent OpenCode installation. The executable,
+  native plugin ID, configuration paths, and ownership markers remain `litopencode`.
+- Migrate published-package and prior package registrations while preserving tuple
+  options, authored routes, and user-owned files. Update checks validate the exact scoped
+  package identity before suggesting or staging a newer stable version.
+- Keep planning agents read-only, permission changes explicit, and workflow evidence local.
+  Ship the reader-facing communication contract and canonical skill routes with compatibility aliases.
+- Align bilingual installation, safety, removal, community, and privacy guidance with the
+  scoped package. Apply the Ignition vector cover and keep installation commands as editable
+  README text. Align the terminal mark with the approved interlocking Ignition symbol
+  and its orange, lime, and ivory palette.
+- Preserve unattended installer behavior and terminal fallbacks. Covers and maintainer-only
+  materials remain outside the npm payload; native runtime and managed skill resources stay packaged.
+
+## 0.2.7 - 2026-09-03
+
+- Add one advisory reader-facing communication contract to the effective OpenCode system,
+  activation, and agent prompts so routine operational metadata stays out of default replies while
+  material failures, risks, and required actions remain visible.
+- Keep `reader`, `technical`, and `audit` request-scoped, and require parent agents to filter child
+  reports without reducing internal DoneClaims, verification, evidence, ledgers, or handoff detail.
+- Add regression and package-parity coverage for prompt injection, compaction, delegation, handoff,
+  structured output, and explicit-detail overrides.
+- Make JSON version lockstep count only the package identity fields, so dependency metadata that
+  happens to match the release version cannot block a valid bump.
+
+## 0.2.6 - 2026-09-02
+
+- Move vendored handoff and scientific-visualization corpora to `vendor/` with
+  thin skill wrappers remaining in the skill paths.
+- Count vendor license, provenance, and NOTICE companions in family
+  payload-parity so packed skills keep the material they name.
+- This is a local release candidate only; no publish, tag, or push was performed.
+
+## 0.2.5 - 2026-09-02
+
+- Hardened inode-reuse checks across skill-loop roots, runtime and anchored-worker file operations,
+  knowledge state, observer logs/locks, and Autoresearch/Autoconference publication. Stable stat
+  identity now carries device/inode plus ctime and birthtime (directory checks use birthtime), with
+  regular-file link-count checks at descriptor and cleanup boundaries.
+- Bound trusted review launchers and shebang interpreters to SHA-256 bytes read through `O_NOFOLLOW`
+  descriptors and verified before and after use; failed observer lock creation now retires and
+  rechecks the lock before removal, and only cooperative parent races are retried.
+- Added eight inode-reuse simulations covering launcher, directory/file, anchored worker, knowledge,
+  lock, observer cleanup, and publication paths; regenerated managed-skill hashes.
+- This is a local release candidate only; no publish, tag, or push was performed.
+
+## 0.2.4 - 2026-09-01
+
+- Added the stacked LitFamily ASCII wordmark to the LitOpenCode CLI and both English and Korean
+  README landing pages; the wordmark remains version-free.
+- Made interactive terminal installs show the model route picker by default, with the current
+  managed lead and helper routes pre-selected. `--yes`, `--no-model-prompt`, dry-run, and
+  non-interactive paths retain their existing behavior.
+- Extended version lockstep coverage to every registered version-bearing file, counting both
+  literal and escaped-regex spellings, and added repository registry tests.
+- Wired the credential-free behavior-replacement, installed-resource-tamper, negative-gate,
+  rules-glob-differential, and Wikify surface QA layers into CI with named exclusions for the
+  composite probe to avoid duplicate work.
+- Kept local OpenCode runtime state out of tracked files. This is a local release candidate only;
+  no publish, tag, or push was performed.
+
+## 0.2.3 - 2026-08-31
+
+- Added packed-payload substance, cross-product parity, and referenced-path checks
+  so installed OpenCode skills retain the material they name.
+- Refreshed the canonical legal attribution metadata without changing the four
+  upstream attributions it records. This is a local release candidate only; no
+  publish, tag, or push was performed.
+
+## 0.2.2 - 2026-08-30
+
+- Added installer model/provider choice with separate lead and helper routes,
+  including `-fast` model-id normalisation so the lead upgrade applies to
+  `lit-loop`, `momus`, and `litwork-reviewer` defaults.
+- Routed the review child through the configured provider credential descriptor,
+  added a symlink-safe host auth-store read, and retained bounded proposal-only
+  review behavior.
+- Kept the plan-file gate and Windows fail-closed learning-loop boundary visible
+  in the OpenCode docs and runtime surfaces.
+- This is a local release candidate only; no publish, tag, or push was performed.
+
+## 0.2.1 - 2026-08-29
+
+- Added the OpenCode-native skill learning loop: user-invoked review or validated proposal import
+  creates pending records, and only an explicit foreground `skill-loop apply` can write an
+  agent-marked skill under the OpenCode user skill root. Shipped, installer-managed, unmarked,
+  symlinked, secret-bearing, and non-portable targets fail closed.
+- Added strict proposal and decision-ledger readers, content-addressed apply/rollback receipts, a
+  usage sidecar, deterministic user-invoked curator transitions, and bounded `tool.execute.after`
+  counter/noticing without hook-triggered model calls or automatic application.
+- Enrolled the expanded Skill Observer and its byte-pinned review contract in the managed install
+  and package surfaces. Prepared package, lockfile, README, release-checklist, and release-fixture
+  version surfaces at `0.2.1`; this entry does not imply publication, tagging, or a remote release.
+- Hardened review-host pinning, bounded current-skill context, transaction recovery, and root-anchored
+  mutations. Review, user-skill mutation, rollback, and curator operations now fail closed on Windows
+  with `WINDOWS_SKILL_LOOP_UNSUPPORTED` until native handle and ACL support is available; passive and
+  project-local proposal surfaces remain available.
+
+## 0.2.0 - 2026-08-27
+
+- Added a standalone output-channel declaration across the 41-skill corpus, with the two
+  activation-byte-budgeted UI/UX contracts declaring from their reachable
+  `references/complete-contract.md` files. A structural documentation gate now rejects missing
+  declarations, unknown artifact genres, and genre/channel mismatches.
+- Added a deliverable hedge guard to OpenCode's `tool.execute.after` surface. That event can inspect
+  bounded text carried by native `write` and `edit` inputs, so the guard appends a
+  model-visible, genre-aware advisory without reopening an artifact path or replacing the successful
+  tool result. Unknown, path-only, and oversized input shapes fail open. The guard reads the active
+  skill's declaration rather than hardcoding an artifact genre.
+- Kept the behavioral claim narrow: two isolated A/B runs found no measurable improvement from the
+  contract prose alone. The declaration is machine-readable metadata consumed by the guard; this
+  release claims that the guard flags qualifying hedges, not that the prose makes the model write a
+  cleaner document.
+- Bumped to `0.2.0` because the corpus-wide contract and new runtime hook are a new product surface,
+  rather than another maintenance increment within the `0.1.x` line.
+
+## 0.1.70 - 2026-08-26
+
+- Aligned the package, lockfile, README surfaces, release checklist, and
+  release fixtures at `0.1.70` for G20 slice 19.
+
+## 0.1.69 - 2026-08-23
+
+- Added the `frontend-ui-ux` taste dials, `browser-drive`, and `skill-observer`
+  workflow surfaces with managed package coverage.
+- Kept package, lockfile, README surfaces, and release-fixture versions aligned at `0.1.69`.
+
+## 0.1.68 - 2026-08-15
+
+- Simplified the English and Korean README surfaces and pinned the packaged
+  cover link to the exact release version.
+- Kept package, lockfile, and release-fixture versions aligned at `0.1.68`.
+
+## 0.1.67 - 2026-08-14
+
+- Added provider-free activation and cache measurement contracts with packed
+  consumer coverage and deterministic lock-race handling.
+
+## 0.1.66 - 2026-08-12
+
+- Added advisory evidence review guidance to the `frontend-ui-ux` skill.
+- Added managed-skill manifest and documentation coverage for the guidance.
+
+## 0.1.65 - 2026-08-11
+
+- Added project-local Wikify knowledge capture, review, save, query, and status flows. Structured
+  events remain `review-needed` until an explicit review changes their state. Queries return only
+  accepted records with bounded deterministic output, provenance, and `claims.jsonl` as the sole authority.
+- Hardened the knowledge store against lock and publication races. Concurrent duplicate writers converge
+  on one record. Lock-owner disappearance, authority replacement, staged writes, atomic renames, and
+  recovery boundaries fail closed without false success results.
+
+## 0.1.64 - 2026-08-09
+
+- Added a default-on foreground update barrier at plugin startup for eligible
+  interactive lifecycle paths. It pins an exact stable package, uses a
+  credential-free npm environment and transaction lock, verifies with doctor,
+  and records rollback/unknown-state receipts before exposing hooks.
+- Kept the detached notifier cache-only and added explicit opt-outs for users
+  who want advisory checks without automatic installation.
+
+## 0.1.63 - 2026-08-06
+
+- Added `outputStyle` config field (`off | asd-ste100 | asd-ste100-ko | eli5 | eli5-ko`) that injects
+  a style instruction into the system prompt after repository rules. The install prompt offers a
+  five-choice menu (default: keep current / off). Validated in the config parser; fails open with no
+  push when the file is missing or the style is `off`.
+
+## 0.1.62 - 2026-08-05
+
+- Re-routed the shipped agent default from `gpt-5.6-sol`/`high` (with `gpt-5.6-terra`/`xhigh` for
+  `lit-explorer` and `lit-librarian`) to `gpt-5.6-luna`/`max` for every shipped agent, following
+  price/performance data showing Luna at max effort matches Sol at high effort on pass rate at
+  roughly a quarter of the cost. The interactive picker now offers only LUNA, and the effort floor
+  moved with it: below-high Luna effort is rejected the same way below-high TERRA effort was
+  before.
+
+## 0.1.61 - 2026-08-04
+
+- Added one exact chat-safe checklist grammar for `lit-plan`: consecutive column-zero rows under
+  `## TODOs` carry concrete Action, Output, and Verification fields, and rows under
+  `## Final verification` carry concrete Verification fields. The planning agent, `/lit-plan`
+  command hook, installed skill, and optional scaffolder expose that shape, while `/start-work`
+  remains separately routed to `lit-implement` and `lit-plan` retains its edit, bash, task, and write
+  denials.
+- Made the optional file-backed checker require active concrete rows, reject placeholder, blank,
+  malformed-marker, misnumbered, and completed-only handoffs, including exact case-insensitive
+  `TBD`/`TODO` and ellipsis-only fields, and exclude fenced, nested, indented, and incidental decoys.
+  `managed-skill-manifest.json` coverage and the focused regression were updated with the contract.
+
+## 0.1.60 - 2026-08-02
+
+- Fixed the `frontend-ui-ux` CLI entrypoint guard, which compared an unresolved `process.argv[1]`
+  against `import.meta.url`. Reached through a symlink — which is how an installed skills root
+  reaches it — the CLI never ran: an invalid design contract and an unknown subcommand both exited 0
+  with empty output, voiding the 0/1/2 contract the file's own header declares. Both sides are now
+  resolved with `realpathSync`.
+- Gave the `frontend-ui-ux` JSON trust layer machine-readable error codes. `strict-json`,
+  `bounded-json`, `stdin-json`, and `uiux` threw bare `Error`/`TypeError`, so a caller could not tell
+  a malformed payload from an unusable file from a bad argument without matching on prose. They now
+  raise a `ContractError` carrying `JSON_INVALID`, `FILE_INVALID`, or `ARGUMENT_INVALID`, with
+  `DESIGN_CONTRACT_INVALID` available to callers. Exit codes are unchanged.
+- Added a language server catalog to `lsp-setup`. The skill's method asks which extension is
+  unserved but shipped nothing that could answer it. `references/` now maps 20 languages to their
+  server, install cost, alternatives, the failure modes that produce confidently wrong diagnostics,
+  and the fallback to use while a file stays unserved.
+- Added runtime, escalation, and tool references to `debugging`: how to obtain a real observation in
+  Node, Python, Go, Rust, a compiled binary without source, and a bundled JS executable; the
+  two-dead-rounds reframe; and the specialist tools for symptoms ordinary instrumentation cannot see.
+- Added per-language references to `programming` for Go, Python, Rust, and TypeScript, covering type
+  patterns, failure handling, concurrency, testing, and strict tooling configuration. These are
+  defaults for new code — the surrounding project's conventions still win.
+
+## 0.1.59 - 2026-08-02
+
+- Added the `lit-comprehend` runtime skill and `/lit-comprehend` command for when understanding,
+  not status, is the bottleneck: after a long agent session it builds one self-contained explainer
+  artifact a person can reason with, rather than the chronology `lit-recap` already provides. It
+  anchors on what the reader already knew, orders the walkthrough conceptually instead of by file,
+  ships an interactive micro-world, discloses what is not verified, and closes with a quiz framed
+  as a speed regulator.
+- Gated execution rather than narrowing activation: being invoked is not permission to build. An
+  invocation naming a path or git range proceeds; a bare invocation or a prose question gets a
+  one-screen scope proposal (target, exclusions, estimate) and waits for approval, with the
+  cheaper one-sentence answer offered when the request deserves it. The proposal is derived from
+  `git status`, `git diff --stat`, and durable state, never by reading the tree.
+- Bundled `scripts/verify-explainer.ts`, an HTML scaffold, and authoring references. The verifier
+  fails an artifact for a phantom code quote, a quote attributed to a missing file, code quoted
+  with no attribution, an external resource reference, a missing canonical section, an artifact
+  written inside the worktree, a quiz option without feedback, a positional tell, collapsed code
+  blocks, and ASCII-art diagrams.
+- Routing accepts `lit-comprehend`, bare `comprehend`, and `lit comprehend`; `comprehension`,
+  `incomprehensible`, `설명해줘`, `이해가 안 돼요`, and `explain this function to me` are proven
+  by negative-control tests not to activate.
+
+## 0.1.58 - 2026-07-30
+
+- Fixed `lit`/`litwork`/`start-work`/`review-work` status reads crashing with a ledger parse error
+  when the workspace's append-only ledger contains legacy events (for example an `event` key
+  instead of `type`). Status now uses a lenient read that skips invalid lines and reports a bounded
+  diagnostic; strict parsing is unchanged for writes, recovery, and the bounded-authority lifecycle.
+- Moved the CI dry-pack report outside the package root so the guard never scans its own report,
+  and rejected non-finite `run_with_deadline.py` seconds as typed CLI input errors.
+
+## 0.1.57 - 2026-07-29
+
+- Added the exact verified 167-file canonical frontend reference library as an inert, separately
+  managed corpus while preserving the existing normalized retrieval dataset.
+- Added one managed Autoresearch family with ten nested modes, one managed Autoconference family with
+  seven nested modes and fail-closed task capability, and one managed Wikify family with five nested
+  modes and five templates.
+- Added hyphen-native family command aliases and bounded leading chat activation without adding tools,
+  agents, MCP servers, daemons, publication actions, or live-profile mutation.
+- Replaced cumulative README release chronology with current supported behavior and replayable gates.
+
+For earlier versions, use the repository tags and commit history together with the release checklist;
+those are authoritative for the exact source and verification state of each release.

package/CODE_OF_CONDUCT.md

@@ -0,0 +1,30 @@
+# Code of Conduct
+
+LitOpenCode welcomes participants across backgrounds and levels of experience.
+This code applies to project issues, pull requests, reviews, and other spaces
+maintained for the project.
+
+Discuss ideas and behavior respectfully. Make technical criticism specific and
+actionable, respect differing experience, and credit others' work. Ask before
+sharing another person's private information. Accept a maintainer's request to
+stop an unproductive or harmful exchange.
+
+Harassment, discriminatory abuse, threats, sexual harassment, doxxing, and repeated
+personal attacks are unacceptable. Do not use project spaces to target someone
+elsewhere. Disagreement about a technical decision is welcome; intimidation is not.
+
+Maintainers may edit or remove harmful content, lock a discussion, warn a
+participant, or restrict participation according to the conduct and its impact.
+They should explain the reason where doing so does not expose private information.
+Reports and responses should disclose only what is needed to address the concern.
+
+To report a concern, use an existing private maintainer contact if one has been
+provided. This repository does not publish a dedicated conduct-reporting address.
+Otherwise, ask for a private contact in a minimal issue without identifying affected
+people or posting sensitive details. Do not use a public issue to disclose private
+evidence. A maintainer involved in a report should defer its handling to another
+available maintainer; where that is not possible, the limitation should be stated.
+
+Participants may ask for a decision to be reconsidered with new context through
+the same reporting channel. No response deadline or confidential service beyond
+the actual reporting channel is promised.

package/CONTRIBUTING.md

@@ -0,0 +1,82 @@
+# Contributing to LitOpenCode
+
+LitOpenCode is an OpenCode plugin with its own CLI, command and skill payloads,
+configuration, and project-local evidence state. Start with the [README](README.md)
+and [workflow reference](docs/reference.md). Contributions are covered by the
+[MIT license](LICENSE), whose holder is LitOpenCode contributors, and the
+[Code of Conduct](CODE_OF_CONDUCT.md).
+
+## Before changing behavior
+
+Open a focused issue for a reproducible bug or a concrete proposal. Describe the
+OpenCode surface involved: installer, doctor, command, chat hook, tool, skill, or
+project ledger. Use [Security](SECURITY.md) for vulnerabilities and
+[Support](SUPPORT.md) for configuration questions. Remove secrets and private
+project content from examples; see [Privacy](docs/privacy.md).
+
+Read the repository's `AGENTS.md` when working with an agent. Keep changes inside
+this product. Other family products are independent implementations, not shared
+runtime dependencies. Preserve unrelated worktree changes and user-owned settings.
+Do not edit the vendored corpus to alter product behavior.
+
+## Local development
+
+Use Node.js 22, matching CI, and npm with the committed lockfile:
+
+```sh
+npm ci
+npm run build
+node --test test/static-workflow-command.test.mjs
+```
+
+Choose the focused test for your change; the command above is an example for
+command behavior. Tests use `node:test`. TypeScript relative imports must retain
+their explicit `.ts` extensions. Reuse existing helpers and native hooks before
+adding a new surface. Add a failing regression for changed behavior, including
+unsafe or invalid input where applicable, then make the smallest sufficient fix.
+
+Run these gates serially before requesting review:
+
+```sh
+npm test
+npm run typecheck
+npm run build
+npm run scan:legacy-tokens
+npm run check:version
+npm run check:pack-payload
+npm run check:managed-skill-manifest
+```
+
+`npm test` includes a build through its pretest hook. The pack guard also builds.
+Do not run concurrent builds in one checkout. Redirect long output to a file and
+record the command's actual exit code; piping into a final-line filter can hide
+failure details. Preserve failed runs when reporting a later fix.
+
+When managed skill files change, regenerate their manifest with
+`npm run gen:managed-skill-manifest` and inspect the resulting diff. Do not update
+hashes merely to conceal an unexpected payload difference. Native CLI, hook,
+ownership, and installed-resource changes need an isolated real-surface probe as
+well as unit tests. Use disposable HOME, OpenCode config, temporary, and npm cache
+roots; never test installation against a contributor's live configuration.
+
+CI also runs behavior replacement, installed-resource tamper, negative-gate,
+rules-glob, and Wikify probes. Its evidence-ledger stress test remains a separate,
+non-blocking diagnostic; report any failure without treating it as a pass. Optional
+scientific Python dependencies can produce an explicit skip or degraded status;
+record that limitation rather than claiming the scientific path was verified.
+
+## Pull requests
+
+Explain the user-visible problem, final behavior, affected surfaces, and validation.
+Include meaningful negative cases and any remaining limits. Keep local evidence,
+ledgers, credentials, temporary installs, and generated tarballs out of the diff.
+Clean up owned probe processes and temporary roots after collecting evidence.
+
+Preserve the planner's denied edit, shell, and delegation permissions. Keep native
+registration IDs, ownership receipts, and config paths stable when changing npm
+identity. A local package probe does not establish public availability or an
+authenticated OpenCode session.
+
+Versions and release actions are maintainer-controlled. A contribution or passing
+CI does not authorize a version change, tag, remote update, or publication. Follow
+the repository-only maintainer release checklist when separately authorized.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 LitOpenCode contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.