@litfamily/litopencode 1.0.0

package/docs/reference.md

@@ -0,0 +1,444 @@
+# LitOpenCode workflow reference
+
+[Quick start](../README.md) · [한국어](reference-Ko-KR.md)
+
+*Workflow discipline in one view: a plan earns approval only with binary checks, and a slice closes only after the real surface, evidence, and cleanup are complete. Tests alone do not close the slice.*
+
+```mermaid
+flowchart TD
+    R["a request<br/>make it better"] --> DI["<b>deep-interview</b><br/>turn it into a decision-complete brief"]
+    DI --> P["<b>lit-plan</b><br/>objective · non-goals<br/>action / output / <b>binary verification</b>"]
+    P --> GATE{"user approves?"}
+    GATE -->|no| P
+    GATE -->|yes| SW["<b>start-work</b><br/>execute one slice"]
+
+    subgraph LOOP["each slice: RED to GREEN to SURFACE to CLEAN"]
+        SW --> RED["failing test first"]
+        RED --> GREEN["smallest change that passes"]
+        GREEN --> SURF["exercise the <b>real surface</b><br/>not just the test"]
+        SURF --> CLEAN["tear down · cleanup receipt"]
+    end
+
+    CLEAN --> EV{"evidence complete?"}
+    EV -->|"tests only"| SW
+    EV -->|"artifact + receipt"| RW["<b>review-work</b><br/>scope · evidence · payload<br/>security · real surface"]
+    RW -->|findings| SW
+    RW -->|clean| HO["<b>lit-handoff</b><br/>resumable packet"]
+
+    style GATE fill:#fff3cd,stroke:#856404
+    style EV fill:#fff3cd,stroke:#856404
+    style SURF fill:#d4edda,stroke:#155724
+    style RW fill:#d1ecf1,stroke:#0c5460
+```
+
+## Install
+
+The command uses an explicit npm package and the preserved executable:
+
+The scoped package is the npm install target. See [package migration](migration.md#scoped-npm-package-migration).
+
+```sh
+npm exec --package @litfamily/litopencode@latest -- litopencode install
+```
+
+Preview the config changes without writing them, or inspect an installed setup:
+
+```sh
+npm exec --package @litfamily/litopencode@latest -- litopencode install --dry-run
+npm exec --package @litfamily/litopencode@latest -- litopencode doctor
+```
+
+The installer registers the plugin through OpenCode's plugin installer. It creates
+`~/.config/opencode/litopencode.json` only when that route file is missing. Existing
+routes remain user-owned. The default OpenCode config is
+`~/.config/opencode/opencode.jsonc`, or `$XDG_CONFIG_HOME/opencode/opencode.jsonc`
+when `XDG_CONFIG_HOME` is set.
+
+*OpenCode provides the event surfaces; LitOpenCode turns them into bounded routes and guards, then records the durable ledger. This is the host connection behind the commands above.*
+
+```mermaid
+flowchart LR
+    subgraph OC["OpenCode"]
+        E1["chat.message"]; E2["command.execute.before"]
+        E3["tool.execute.before"]; E4["tool.execute.after"]; E5["event"]
+    end
+    subgraph LO["LitOpenCode plugin"]
+        RT["TypeScript runtime<br/>routing · bounded authority"]
+        GUARD["tool guards"]
+        LEDGER[("durable ledger")]
+    end
+    E1 --> RT; E2 --> RT; E3 --> GUARD; E4 --> GUARD; E5 --> RT
+    RT --> LEDGER; GUARD --> LEDGER
+    LO --> S["41 skills"]
+```
+
+Optional permission and model choices are explicit:
+
+```sh
+npm exec --package @litfamily/litopencode@latest -- litopencode install --permission-prompt
+npm exec --package @litfamily/litopencode@latest -- litopencode install --permission-mode balanced
+npm exec --package @litfamily/litopencode@latest -- litopencode install --yolo
+npm exec --package @litfamily/litopencode@latest -- litopencode install --provider openai --model gpt-6-astra --effort xhigh
+npm exec --package @litfamily/litopencode@latest -- litopencode install --provider openai --model gpt-5.6-luna --effort max
+npm exec --package @litfamily/litopencode@latest -- litopencode install --provider xai --model grok-4.6 --effort xhigh
+npm exec --package @litfamily/litopencode@latest -- litopencode install --model-prompt
+```
+
+The interactive picker (`--model-prompt`, or a fresh TTY install) asks for a provider
+(OpenAI GPT-6 Astra / GPT-5.6 family or xAI Grok), a LEAD model for the planning/review categories, and a
+HELPER model for the execution/research categories; the helper may use a different provider.
+OpenAI rows are gpt-6-astra · xhigh (low, medium, high, xhigh, or max), gpt-5.6-sol · xhigh,
+gpt-5.6-luna · max, gpt-5.6 · high, and gpt-5.6-terra · xhigh; xai rows are the
+reasoning-capable Grok models from the host catalog, `grok-4.6` first.
+`--subagent-model <provider/id>` and `--subagent-effort <e>` set the helper route from flags.
+Legacy OpenAI `<id>-fast` aliases remain accepted and policy-checked as their base ids;
+`gpt-6-astra-fast` is not invented or accepted. Choosing Grok
+without an xAI credential prints a warning naming `XAI_API_KEY` and still writes the route.
+Existing non-default routes are preserved by default (`--no-model-prompt` and `--yes` also
+preserve); `--model-prompt` rewrites managed keys only.
+
+`install --yes` skips every picker and keeps the saved output style. CI and non-TTY
+installs skip automatic prompts; explicit prompt flags still work. `NO_COLOR` keeps
+interactive style selection available without color or cursor escapes. Dumb terminals
+and non-UTF-8 locales use the plain `LIT` wordmark without ANSI decoration.
+
+The OpenCode config hook routes the lead categories and `lit-loop` to
+`openai/gpt-6-astra` with native `reasoningEffort` (fresh/reset default `xhigh`).
+Execution and research helpers retain `openai/gpt-5.6-luna`/`max` unless a helper route is
+explicitly selected. The hook registers only missing native Astra model metadata
+(`reasoning`, no `temperature`, and tool calls); existing provider/model fields and host
+agent overrides are preserved. `momus` and `litwork-reviewer` are never force-overwritten.
+LUNA plus `xhigh` is rejected, while Astra accepts all five supported efforts.
+
+The default is safe/ask-first behavior. `balanced` and `yolo` are opt-in. A global
+installation is also supported:
+
+```sh
+npm install -g @litfamily/litopencode
+litopencode install
+```
+
+## First use
+
+Restart OpenCode after installation and press Tab. The visible agent switcher should contain:
+
+| Agent | Use it for |
+| --- | --- |
+| `lit-loop` | Implement, verify, delegate safely, and record durable progress. |
+| `lit-plan` | Produce an evidence-backed plan. `edit`, `bash`, and `task` stay denied. |
+| `lit-implement` | Execute an approved plan after `/start-work`. |
+
+The normal workflow is:
+
+1. Ask `lit-plan` for one bounded, verifiable plan.
+2. Approve it, then run `/start-work`.
+3. Run `/review-work` before making a completion claim.
+
+The OpenCode-native workflow contract keeps practical delivery discipline: plans are
+objective-achievable, use adaptive detail, and allow maximum safe subagent delegation.
+`lit-plan` waits for explicit user confirmation; `/start-work` is an execution-only handoff
+for an approved plan; `/review-work` checks the `DoneClaim` through a five-lane review.
+
+## Core commands
+
+| Route | Purpose |
+| --- | --- |
+| `/lit` | Start the normal `lit-loop` workflow. |
+| `/litwork` | Enter the bounded work loop. |
+| `/lit-plan` | Create a read-only implementation plan. |
+| `/start-work` | Execute an approved plan through `lit-implement`. |
+| `/review-work` | Review a draft plan or completed work. |
+| `/litresearch` or `/lit-research` | Run evidence-backed research with a sequential fallback. |
+| `/lit-handoff` | Create a resumable handoff; exact bare `handoff` also routes here. |
+| `/lit-recap` | Read a concise recap from local ledger and session context. |
+| `/lit-scientific-visualization` | Use the packaged scientific-visualization workflow. |
+| `/lit-korean` | Review prose for natural phrasing without changing meaning. |
+| `/text-neutralization` | Review prose with the neutralization contract. |
+
+Additional static routes include `/refactor`, `/lit-burnoff`, `/lit-code`,
+`/debugging`, `/lit-commit`, `/lsp`, `/lsp-setup`, `/rules`, `/deep-interview`,
+`/structural-search`, the `/autoresearch` and `/autoconference` families, and the
+`/wikify-*` routes. The Autoresearch, Autoconference, and Wikify families add guidance or
+bounded local behavior; they do not silently grant new host authority. Frontend UI/UX also
+ships a canonical frontend library for its offline, read-only design references.
+
+The renamed routes are `/lit-crucible`, `/lit-init`, `/lit-commit`, `/lit-burnoff`,
+`/lit-burnoff-file`, `/lit-korean`, `/lit-fetch`, and `/lit-code`. Previous names remain
+compatibility redirects for one release and print one deprecation note. Install/update
+removes installer-owned old skill directories after verifying their replacements; user-owned
+files are preserved. See [migration notes](migration.md#skill-id-renames).
+
+### Durable loop state
+
+LitOpenCode owns durable goal state under `.litopencode/litgoal/lit-loop/`. The append-only
+ledger records progress and evidence. The CLI can inspect and update that ledger:
+
+```sh
+litopencode status
+litopencode create-goals --session-id <id> --objective <text>
+litopencode record-evidence --criterion-id <id> --kind <red|green|scenario|cleanup|note> --ref <ref>
+litopencode checkpoint --summary <text>
+litopencode complete-goals
+```
+
+`/start-work init|resume|cancel|complete|status <schema-3 JSON>` is the bounded-authority
+surface for approved execution. Resume requires the trusted user route and matching work,
+session, revision, and grant; copied or quoted text is inert.
+
+Native goal verdict: the OpenCode host surface does not currently expose a native goal primitive;
+LitOpenCode uses `.litopencode/litgoal` instead. Static skills are installed under
+`skills/*/SKILL.md` and remain documentation unless a documented OpenCode route selects them.
+
+## Skill learning loop
+
+Skill Observer can turn bounded observations and a per-session tool-iteration reminder into
+pending skill proposals. The hook only updates project-local counters and surfaces a notice; it
+does not call a model or edit a skill. Review remains a foreground user action:
+
+```sh
+litopencode skill-loop review
+litopencode skill-loop propose <proposal.json>
+litopencode skill-loop list
+litopencode skill-loop apply <proposal-id>
+litopencode skill-loop reject <proposal-id>
+litopencode skill-loop rollback <ledger-id>
+litopencode skill-curator status
+litopencode skill-curator run
+```
+
+Nothing is applied without the explicit `skill-loop apply` command; shipped configuration keeps
+`autoApply: false`. New skills are written only to the OpenCode user skill root —
+`$XDG_CONFIG_HOME/opencode/skills/<name>/`, or `~/.config/opencode/skills/<name>/` when XDG is
+unset — with the distinct `metadata.litopencodeAgentGenerated: "true"` marker. Shipped,
+installer-managed, unmarked, and unsafe targets are refused.
+
+`skill-loop review` runs in an empty temporary home/config directory with tools and permissions
+denied, on the configured `review` category route (provider, model, and effort from
+`litopencode.json`) so a Grok-configured install never needs an OpenAI token. Set
+`LITOPENCODE_REVIEW_HOST` only to an absolute OpenCode launcher you intend LitOpenCode
+to pin and execute. If a host-native credential is already available as a narrow environment
+variable, opt in by naming it with `LITOPENCODE_REVIEW_CREDENTIAL_ENV=NAME[,NAME...]`; when that
+descriptor is unset, the configured review provider's own variable (for example `XAI_API_KEY`)
+is forwarded if present. OAuth-only hosts can opt in with `LITOPENCODE_REVIEW_AUTH=host`, which
+copies exactly the configured review provider's entry from the OpenCode auth store into the
+isolated child — nothing else. Help probes remain credential-free, and credential values are
+never placed in arguments, prompts, logs, or evidence.
+
+The learning loop's review, user-skill mutation, rollback, and curator boundaries currently support
+POSIX hosts (including macOS and Linux). On Windows those global/model boundaries fail closed with
+`WINDOWS_SKILL_LOOP_UNSUPPORTED` pending native handle and ACL support. Project-local listing and
+proposal intake remain available; this limitation does not affect install, doctor, or other passive
+LitOpenCode surfaces.
+
+Each successful mutation records content-addressed before/after state in the project-local
+`.litopencode/skill-ledger.jsonl`; rollback first verifies that current bytes still match the
+recorded after-image. The shipped config records a seven-day curator interval and a two-hour idle
+window for callers, but this release starts no background scheduler. A user-invoked
+`skill-curator run` marks eligible active entries stale after 30 days and moves already-stale
+entries to a recoverable archive after 90 days. Pinned entries are skipped, and archive transitions
+preserve bytes rather than deleting them.
+
+## Runtime Skills
+
+*What matters at install time is the packed payload, not a source-tree listing: every declared skill must resolve its own references before it can be trusted on a fresh machine.*
+
+```mermaid
+flowchart LR
+    SK["a skill"] --> Q{"does it declare<br/>a capability?"}
+    Q -->|"self-contained<br/>procedure"| AL["explicit allowlist entry<br/>with a written reason"]
+    Q -->|"needs a corpus"| C["corpus must resolve<br/>inside the <b>packed payload</b>"]
+    AL --> G1
+    C --> G1["<b>payload-substance</b>"]
+    G1 --> G2["<b>cross-product parity</b><br/>one product cannot ship a stub<br/>where the family ships substance"]
+    G2 --> G3["<b>referenced-path resolution</b><br/>every path in a SKILL.md<br/>must exist in the tarball"]
+    G3 --> OK["installs and works<br/>on a machine that has<br/>nothing else"]
+    style C fill:#d4edda,stroke:#155724
+    style OK fill:#d4edda,stroke:#155724
+```
+
+<details>
+<summary>Static skills available for native discovery</summary>
+
+- <code>agent-roster</code>
+- <code>lit-burnoff-file</code>
+- <code>autoconference</code>
+- <code>autoresearch</code>
+- <code>browser-drive</code>
+- <code>comment-checker</code>
+- <code>debugging</code>
+- <code>deep-interview</code>
+- <code>doctor-installer</code>
+- <code>durable-litgoal</code>
+- <code>frontend-ui-ux</code>
+- <code>lit-commit</code>
+- <code>lit-crucible</code>
+- <code>lit-init</code>
+- <code>lit-comprehend</code>
+- <code>lit-handoff</code>
+- <code>lit-plan</code>
+- <code>lit-recap</code>
+- <code>lit-scientific-visualization</code>
+- <code>litresearch</code>
+- <code>litwork</code>
+- <code>lsp</code>
+- <code>lsp-setup</code>
+- <code>native-goal-verdict</code>
+- <code>lit-code</code>
+- <code>lit-fetch</code>
+- <code>refactor</code>
+- <code>reference-benchmark-claims</code>
+- <code>release-guardrails</code>
+- <code>lit-burnoff</code>
+- <code>review-work</code>
+- <code>rules</code>
+- <code>search-workflow-ideas</code>
+- <code>skill-observer</code>
+- <code>start-work</code>
+- <code>structural-search</code>
+- <code>lit-korean</code>
+- <code>tool-guards</code>
+- <code>visual-qa</code>
+- <code>wikify</code>
+- <code>workflow-loop</code>
+
+</details>
+
+### UI and Visual QA managed helpers
+
+The `frontend-ui-ux` and `visual-qa` routes use Node ESM helpers and bounded,
+read-only evidence. They do not add workflow-family scaffold or status helpers.
+
+### Managed workflow families
+
+Autoresearch and Autoconference provide the scaffold/status helpers for their
+bounded local workflows. Wikify owns templates and contracts, but provides no scaffold or status
+helper. Wikify probe checks five surfaces: `tool.execute.after`,
+`chat.message`, `command.execute.before`, and the remaining host hook surfaces
+listed in the Wikify contract. Wikify additionally owns one bounded `wikify`
+tool on `tool.execute.after`; it does not expose Autoresearch or Autoconference
+tools.
+
+## Release Readiness
+
+Use [`release-checklist.md`](release-checklist.md) for maintainer-only
+release gates; this README documents the current installed behavior.
+
+## Verify
+
+From a LitOpenCode checkout, run the focused package and source gates:
+
+```sh
+npm install
+npm run build
+npm test
+npm run typecheck
+npm run check:managed-skill-manifest
+npm run scan:legacy-tokens
+npm run check:version
+npm run check:pack-payload
+npm run qa:real-surface
+npm pack --dry-run --json --ignore-scripts
+```
+
+The no-write local probes are:
+
+```sh
+node bin/litopencode.cjs doctor --root .
+node bin/litopencode.cjs install --dry-run --root .
+```
+
+`doctor` reports package, route, command, skill, and local-state health without writing.
+The pack check proves that the compiled plugin, CLI, types, public skills, README, and
+required docs are present while local state and source-only files stay out of the payload.
+These checks are reproducible from a checkout.
+
+## Remove
+
+There is no `uninstall` subcommand. For a global install, remove the npm binary first:
+
+```sh
+npm uninstall -g @litfamily/litopencode
+```
+
+Then remove the `@litfamily/litopencode` or `@litfamily/litopencode@<version>` entry from the `plugin` array in
+OpenCode's managed `opencode.jsonc` (or the custom root's `opencode.json`). Remove
+`~/.config/opencode/litopencode.json` only when you no longer need its user-owned routes.
+Project `.litopencode/` state is separate; preserve it for later resume or remove it only
+when discarding that project's ledger and receipts is intentional.
+
+## Safety and updates
+
+- `lit-plan` remains planning-only: its `edit`, `bash`, and `task` permissions stay denied.
+- `balanced` and `yolo` are opt-in permission modes. Dangerous shell patterns remain on ask in
+  balanced mode; YOLO allows the configured OpenCode tools, but it does not weaken the planner's
+  deny guard or allow recursive subagent delegation.
+- Packaged startup and successful interactive `install`/`doctor` commands use a bounded,
+  exact-version foreground update barrier. Disable it with `--no-auto-update` or
+  `LITOPENCODE_NO_AUTO_UPDATE=1`. The cache-only notice can also be disabled with
+  `NO_UPDATE_NOTIFIER=1` or `LITOPENCODE_NO_UPDATE_CHECK=1`.
+- Public-source retrieval uses SSRF checks, redirect validation, byte limits, and access
+  verdicts. Fetched text is treated as data, not as instructions.
+
+- repository rules engine with two lanes:
+  The static lane has a final encoded per-rule limit of 12,000 characters; static total remains 40,000.
+  The dynamic lane has a final encoded per-rule limit of 4,000 characters; its total limit of 10,000 characters.
+  Both retain complete rule fragments, XML entities, and code points.
+  Metadata-only overflow emits a bounded digest/omission rule fragment.
+- session-scoped rule delivery deduplicates matching content and stays bounded after compaction.
+
+## Package surface
+
+- Package and plugin id: `litopencode`
+- CLI binary: `litopencode`
+- Public exports: `litopencode` and `litopencode/server`
+- Global route file: `~/.config/opencode/litopencode.json`
+- Project overrides: `.litopencode/config.json`
+- Durable state: `.litopencode/litgoal/`
+- User state root: `~/.litopencode/`
+
+## Documentation
+
+- [Quick start](../README.md)
+- [한국어 상세 안내](reference-Ko-KR.md)
+- [Changelog](../CHANGELOG.md)
+- [Migration notes](migration.md)
+- [Terminal mark and activation probes](lit-mark.md)
+- [Maintainer release checklist](release-checklist.md)
+
+
+## Installed canonical skill assets
+
+`lit-handoff` and `lit-scientific-visualization` install their exact authored source
+inside `skills/<id>/canonical/` below the selected OpenCode config root. Read the
+native skill's `SKILL.md` first and resolve its `exact_source_root` from that file's
+directory. Slash and bare-route prompts use the same native skill location; the
+project directory and command directory are not reference roots. A packed install
+carries the full source closure and does not require the developer's checkout.
+
+Doctor checks the installed file inventory and hashes as well as package source.
+Missing Python or matplotlib is a separate `DEGRADED` capability result; missing
+installed skill files are an integrity error. Neither check installs dependencies.
+
+Repeat installation preserves an unmarked user-owned skill. For a managed wrapper,
+the new canonical subtree must match its exact packaged files and hashes before
+replacement; modified, foreign, missing, or symlinked content there causes refusal.
+Back up your changes outside the managed tree before retrying. A previous flat
+managed wrapper without this subtree can gain the missing assets. Existing generated
+wrapper and distribution files remain installer-managed and can be replaced; this
+is not a promise to preserve edits to every generated file.
+
+
+Ordinary Python imports may create `__pycache__` files inside the scientific skill.
+Doctor lists eligible paths in `canonicalCacheFiles`; it checks that each regular
+CPython/PyPy tagged `.pyc` (including optimized tags) maps to an unchanged owned `.py`
+source. It does not load or trust the bytecode contents. Other extras, missing or
+modified sources, and symlinked files/directories still fail integrity checks.
+
+On a cache-bearing repeat install, the previous whole skill tree is retained under
+`<config-root>/.litopencode-canonical-backups/<unique-id>/skill`, outside native skill
+discovery. `prepared.json` records the planned move; `receipt.json` records completed
+retention and each cache SHA-256. A fresh canonical copy becomes active, so the retained
+bytecode cannot override its source. Cacheless repeat and migration from a flat wrapper
+need no retained backup. If replacement or safe rollback fails, preserve the reported
+recovery paths and inspect them; do not assume installation succeeded. These checks
+cover ordinary use and detected path changes, not hostile concurrent filesystem mutation.

package/output-styles/asd-ste100-ko.md

@@ -0,0 +1,41 @@
+<!-- Principle-adapted from the English version, not translated. Do not re-translate. -->
+
+설명, 요약, 사용자에게 주는 지시는 모두 쉬운 기술 문장체로 쓰십시오. 코드, 코드 주석, 파일 내용,
+인용문에는 이 규칙을 적용하지 마십시오. 그것들은 각자의 규칙을 따릅니다.
+
+## 문장 구성
+
+- 한 문장에 한 가지 생각만 담으십시오. 관계없는 내용을 "그리고"나 "그런데"로 잇지 마십시오.
+- 문장을 짧게 쓰십시오. 설명문은 40자 안팎, 지시문은 30자 안팎을 넘기지 마십시오.
+- 능동태를 쓰십시오. 행동의 주체를 밝히십시오.
+  - "마이그레이션 스크립트를 실행하십시오."라고 쓰십시오. "마이그레이션 스크립트가 실행되어야
+    합니다."라고 쓰지 마십시오.
+- 시제는 단순하게 쓰십시오.
+  - 사실이나 일반 원리는 현재형: "이 함수는 null을 반환합니다."
+  - 끝난 일은 과거형: "빌드가 실패했습니다."
+  - 앞으로 일어날 일은 미래형: "이 테스트는 실패할 것입니다."
+  - "~하고 있었습니다", "~해 왔습니다" 같은 복잡한 시제는 피하십시오.
+- 긴 명사 나열을 피하십시오. 조사로 끊어 쓰십시오.
+  - "연결 풀의 타임아웃 값"이라고 쓰십시오. "연결 풀 타임아웃 값"처럼 명사를 겹쳐 쓰지 마십시오.
+
+## 어휘
+
+- 한 단어에 한 뜻만 쓰십시오. 같은 개념을 한 답변 안에서 다른 말로 바꿔 쓰지 마십시오. ("삭제"면
+  "삭제"로 끝까지 쓰고, "제거"와 섞지 마십시오.)
+- 전문 용어, 관용구, 비유적 표현을 피하십시오. 있는 그대로 말하십시오.
+- 전문 용어는 처음 쓸 때 쉬운 말로 뜻을 설명하십시오.
+- 어렵고 추상적인 말보다 쉽고 구체적인 말을 쓰십시오.
+
+## 지시와 절차
+
+- 지시는 번호가 붙은 단계로, 명령형·현재형으로 쓰십시오.
+- 한 단계에 한 가지 행동만 넣으십시오. 여러 행동은 단계를 나누십시오.
+- 조건을 행동보다 먼저 쓰십시오. "파일이 있으면 삭제하십시오."라고 쓰십시오. "파일을 삭제하십시오,
+  있다면."이라고 쓰지 마십시오.
+- 위험하거나 주의가 필요한 단계 앞에는 `경고:`, `주의:`, `참고:`를 명확히 표시하십시오.
+
+## 구조
+
+- 빽빽한 글보다 짧은 문단, 번호 목록, 글머리표를 쓰십시오.
+- 핵심을 먼저 말하고, 세부 내용을 뒤에 쓰십시오.
+- 위 규칙과 간결함이 부딪히면, 명확하고 오해 없는 뜻을 선택하십시오.

package/output-styles/asd-ste100.md

@@ -0,0 +1,40 @@
+Write all prose output in ASD-STE100 Simplified Technical English (STE). This applies to
+explanations, summaries, and instructions you give the user — not to code, code comments, file
+contents, or quoted material, which follow their own conventions.
+
+## Sentence construction
+
+- One idea per sentence. Do not join unrelated ideas with "and" or "which."
+- Keep sentences short: 20 words or fewer for descriptive sentences, 15-20 words for instructions.
+- Use active voice. Name the agent that does the action.
+  - Write "Run the migration script." not "The migration script should be run."
+- Use simple verb tenses only:
+  - Simple present for facts and general truths ("The function returns null.")
+  - Simple past for completed actions ("The build failed.")
+  - Simple future for future events ("The test will fail.")
+  - Avoid continuous, perfect, and perfect-continuous tenses ("has been running," "will have completed").
+- Always use articles ("a," "an," "the"). Do not drop them for brevity.
+- Limit noun clusters to 3 nouns in a row. Break up long noun strings with prepositions.
+  - Write "the timeout value for the connection pool" not "the connection pool timeout value."
+- Avoid "-ing" forms as nouns or adjectives where a plain verb form exists.
+  - Write "to configure the server" not "for configuring the server."
+
+## Vocabulary
+
+- One word, one meaning. Do not switch between synonyms for the same concept in one response (pick "delete" or "remove," not both).
+- Avoid jargon, idioms, and figurative language. State things literally.
+- Define a technical term the first time you use it, in simple terms.
+- Prefer common, concrete words over abstract or formal ones.
+
+## Instructions and procedures
+
+- Write instructions as numbered steps, in imperative mood, present tense.
+- One action per step. Split compound actions into separate steps.
+- Put a step's condition before its action: "If the file exists, delete it." not "Delete the file if it exists."
+- Flag risk or required attention with an explicit label before the relevant step: `WARNING:`, `CAUTION:`, or `NOTE:`.
+
+## Structure
+
+- Prefer short paragraphs, numbered steps, and bullet lists over dense prose.
+- State the main point first, then supporting detail.
+- If a rule above conflicts with brevity, choose clarity and unambiguous meaning.

package/output-styles/eli5-ko.md

@@ -0,0 +1,13 @@
+<!-- Principle-adapted from the English version, not translated. Do not re-translate. -->
+
+오늘 하루가 길었고 머리가 지쳤어요. 다섯 살짜리한테 설명하듯 말해주세요.
+
+짧은 단어, 짧은 문장, 짧은 문단을 쓰세요. 어려운 단어를 써야 한다면 바로 뒤에 쉽게 풀어서
+설명하세요. 정말 필요한 내용만 답하세요.
+
+무엇을 했는지, 잘 됐는지, 다음에 뭘 해야 하는지만 말해주세요.
+
+제가 뭔가 결정해야 한다면: 선택지는 최대 2개, 빠르게 고르는 데 필요한 맥락만, 그리고 당신이라면
+어떤 걸 고를지도 알려주세요.
+
+경로와 명령어는 정확하게 그대로 써주세요. 그 외의 것까지 신경 쓸 여유가 없어요.

package/output-styles/eli5.md

@@ -0,0 +1,11 @@
+It's been a long day and my brain is fried, talk to me like I'm 5.
+
+Small words, short sentences, short paragraphs. If you have to use
+a big word, explain it right after. Only return what's actually necessary.
+
+Just tell me what you did, did it work, what do I do now.
+
+If I have to decide something: 2 options max, the context I need to pick fast,
+and which one you'd go with.
+
+Keep paths and commands exact. I have no brain cells left for the rest.

package/package.json

@@ -0,0 +1,57 @@
+{
+  "name": "@litfamily/litopencode",
+  "version": "1.0.0",
+  "description": "LitOpenCode OpenCode plugin bootstrap package.",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/wjgoarxiv/litopencode.git"
+  },
+  "homepage": "https://github.com/wjgoarxiv/litopencode#readme",
+  "author": "LitOpenCode contributors",
+  "type": "module",
+  "bin": {
+    "litopencode": "bin/litopencode.cjs"
+  },
+  "types": "./dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js"
+    },
+    "./server": {
+      "types": "./dist/server.d.ts",
+      "import": "./dist/server.js"
+    }
+  },
+  "scripts": {
+    "build": "node tools/run-build.mjs",
+    "check:pack-payload": "npm run build && npm pack --dry-run --json --ignore-scripts | node tools/check-pack-payload.mjs --stdin",
+    "check:version": "node tools/check-version-lockstep.mjs",
+    "check:managed-skill-manifest": "node tools/gen-managed-skill-manifest.mjs --check",
+    "gen:managed-skill-manifest": "node tools/gen-managed-skill-manifest.mjs",
+    "prepack": "npm run build",
+    "pretest": "npm run build",
+    "qa:behavior-replacement": "node tools/run-behavior-replacement-probes.mjs",
+    "qa:installed-resource-tamper": "node tools/run-installed-resource-tamper-probe.mjs",
+    "qa:negative-gate-matrix": "node tools/run-negative-gate-matrix.mjs",
+    "qa:rules-glob-differential": "node tools/run-rules-glob-differential.mjs --check",
+    "qa:real-surface": "npm run build && node tools/run-negative-gate-matrix.mjs && node tools/run-installed-resource-tamper-probe.mjs && node tools/run-behavior-replacement-probes.mjs && node tools/run-wikify-surface-probe.mjs",
+    "scan:legacy-tokens": "node tools/scan-legacy-tokens.mjs",
+    "test": "node --test --test-concurrency=4 \"test/*.test.mjs\"",
+    "test:stress": "node --test test/performance/evidence-ledger-stress.test.mjs",
+    "typecheck": "node tools/run-typecheck.mjs"
+  },
+  "dependencies": {
+    "@opencode-ai/plugin": "^1.17.6",
+    "yaml": "^2.9.0"
+  },
+  "devDependencies": {
+    "@types/node": "^24.12.2",
+    "picomatch": "^4.0.5",
+    "typescript": "^5.9.3"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

package/qa/fixtures/litfamily-harness-speed-v1.json

@@ -0,0 +1,38 @@
+{
+  "schema": "litfamily.harness-speed-scenario/v1",
+  "scenario_id": "litfamily-speed-lit-activation-v1",
+  "encoding": "UTF-8",
+  "prompt_trailing_newline": false,
+  "records": [
+    {
+      "id": "B0",
+      "measured": false,
+      "prompt": "Reply with exactly this ASCII token and nothing else: LIT_SPEED_B0_OK",
+      "prompt_bytes": 69,
+      "prompt_sha256": "a5f634868cc6e20375f60803dcbe850eeb3d2709d6bb2c0efee204b9b5534904",
+      "sentinel": "LIT_SPEED_B0_OK",
+      "sentinel_bytes": 15,
+      "sentinel_sha256": "7652e386a8825b8459833ec8aa7d19fa6a55bf82652f63f45cc4ae29bcfee17b"
+    },
+    {
+      "id": "S1",
+      "measured": true,
+      "prompt": "lit\nReply with exactly this ASCII token and nothing else: LIT_SPEED_S1_OK",
+      "prompt_bytes": 73,
+      "prompt_sha256": "a52e0ec9444cc2195c1c8dbb638862251f1f07df0512c505d06b43d44d4a4e78",
+      "sentinel": "LIT_SPEED_S1_OK",
+      "sentinel_bytes": 15,
+      "sentinel_sha256": "8a655cd5ccd93b2727d641b5ca444e0f72bc09016115b7b6b16aa29e8457414a"
+    },
+    {
+      "id": "S2",
+      "measured": true,
+      "prompt": "Continue the active workflow. Reply with exactly this ASCII token and nothing else: LIT_SPEED_S2_OK",
+      "prompt_bytes": 99,
+      "prompt_sha256": "7b6712f5542798968a74a56241689f40475d50851d5a4d80f877d9a2589983a0",
+      "sentinel": "LIT_SPEED_S2_OK",
+      "sentinel_bytes": 15,
+      "sentinel_sha256": "482d6b2f6dc3a90d5feb4aef7615f715ef2ffbceea5e5ffde90f070175c22dc8"
+    }
+  ]
+}