dsh-standards 1.2.1 → 1.2.2

package/README.en.md

@@ -0,0 +1,347 @@
+<div align="center">
+
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="docs/assets/dsh-standards-flow-dark.svg">
+  <img src="docs/assets/dsh-standards-flow-light.svg" width="100%" alt="dsh-standards pipeline: plugin dir → dsh-standards (DSC-01…13 static) → L1–L3 runtime → report (text/json/sarif, exit 0|1|2|3) → dsh-levels (ENTRY / FLOW / BOOST)">
+</picture>
+
+# dsh-standards
+
+**Find out a DSH plugin's compatibility, usability, and efficiency evidence in one command — static never poses as runtime, and `not-run` never counts as PASS.**
+
+The execution home for the three DSH adoption levels (ENTRY / FLOW / BOOST): the checker, the level audit, and the canonical specification text all live here; the prose guide is [dsh-plugin-guide](https://github.com/PerryLink/dsh-plugin-guide) and the low-level clause text stays in [dsh-plugin-spec](https://github.com/PerryLink/dsh-plugin-spec).
+
+[Install](#three-step-setup) · [Configuration](#configuration) · [FAQ](#faq) · [简体中文](README.md) · [English](README.en.md) · [Español](README.es.md) · [Português](README.pt.md) · [हिन्दी](README.hi.md)
+
+[![License](https://img.shields.io/badge/license-Apache--2.0%20%2B%20CC--BY--4.0-3366CC?style=flat-square&labelColor=2B3245)](LICENSE)
+[![npm](https://img.shields.io/npm/v/dsh-standards?style=flat-square&labelColor=2B3245&color=CB3837)](https://www.npmjs.com/package/dsh-standards)
+[![Node](https://img.shields.io/badge/node-%3E%3D24.9.0-339933?style=flat-square&labelColor=2B3245)](#three-step-setup)
+
+[![Stars](https://img.shields.io/github/stars/PerryLink/dsh-standards?style=flat-square&labelColor=2B3245)](https://github.com/PerryLink/dsh-standards/stargazers)
+[![npm downloads](https://img.shields.io/npm/dm/dsh-standards?style=flat-square&labelColor=2B3245&color=CB3837)](https://www.npmjs.com/package/dsh-standards)
+[![Gitee](https://img.shields.io/badge/Gitee-mirror-C71D23?style=flat-square&labelColor=2B3245&logo=gitee)](https://gitee.com/perrylink/dsh-standards)
+
+```sh
+npm i -g dsh-standards                                        # global: dsh-standards + dsh-levels
+npx dsh-standards <plugin-dir> --scope entry --entry-claim static   # or run it without installing
+```
+
+</div>
+
+<sub>Community project, not affiliated with DeepSeek: non-official, not a certification, not a security audit, no official endorsement. `DeepSeek Harness` is a DeepSeek trademark; this project only uses the `DSH` abbreviation. Canonical repository: <https://github.com/PerryLink/dsh-standards> (Gitee is a mirror).</sub>
+
+## What it is
+
+This repository does not ship plugins. It builds the ruler, then measures with it.
+
+DSH adoption depth has three levels, each with its own question and evidence vocabulary, and they are **never collapsed into one score**:
+
+| Layer | Specification / practice | Question it answers | Current state |
+|---|---|---|---|
+| 1 | **DSH-ENTRY basics** | Can an existing plugin be recognized, installed, and booted without changing its code? | 18/18 `recognized`; base-only 10/18 carry runtime evidence (8 clean + 2 reject); real web profile 16/18 warning-free with 2 host-gate rejects (self-assessed, no independent review); client 15/15 loaded; 18/18 clean not achieved |
+| 2 | **DSH-FLOW smoothness** | Can a user install, configure, use, recover, upgrade, and uninstall the plugin without surprises? | draft; J1/J2/J4/J6 **12 PASS** on fixtures; J3/J7 fixture runtime; J5/J8 static preflight; 10 real packages **78 PASS / 4 FAIL / 2 blocked / 66 not-run** |
+| 3 | **DSH-BOOST acceleration** | Does the plugin use Cordis primitives (effect, inject, volatile, events, seams, jobs, bounded parallelism) to make development and use measurably cheaper? | draft; LEV-1–8 **8/8 measured** (loader / context / provider-capture / synthetic timer; not a production baseline) |
+
+Three things live here: `40-checker/` (the DS-C conformance checker plus every evidence script), `bin/` (two public entry points, `dsh-standards` and `dsh-levels`), and `spec/` with the canonical clause text in the numbered `31`–`35` documents. The tools only read the target package and never modify what they inspect; the verdict only accepts evidence, so `not-run`, `NA`, and `static` are never written up as PASS.
+
+The old npm reservation `dsh-standards@0.0.3` only held the name; this 1.x line replaces it (currently `1.1.2`), and the two bins are `dsh-standards` and `dsh-levels`.
+
+Fixed anchor, recorded in `spec/levels.json` and in every report: `repo=deepseek-ai/deepseek-harness`, `commit=ce7073d51254`, `dsh=0.2.2-alpha.2`, `cordis=4.0.6-alpha.1`. Pass a different runtime version explicitly and the report records which ruler you measured with.
+
+### Current hard numbers
+
+- 18 packages, entry static: **18/18 recognized, 0 clean**;
+- 18 packages, DS-C static: **0 PASS / 9 WARN / 9 FAIL**;
+- 18 packages, base-only runtime: **18/18 raw JSON present; 10/18 G3 verified, 8 partial, 0 error**; after the host gate **8 clean + 2 reject + 8 partial**;
+- client runtime: **15/15 `client-module-loaded`** (real Chromium); web-app **4/4 loaded**;
+- unusable: **0 plugin code FAIL; 2 host-gate rejects (telemetry/verification)**;
+- checker self-test: **128/128 tests** (locally via `node --test 40-checker/test/*.test.mjs`; includes governance, DCO, attestation, claim, verification, chain, and drift tests);
+- J1: **6 PASS + 1 WARN**; LEV-3: volatile **0/20** vs control **20/20** remount; LEV-6: stable prefix **100%** vs volatile **0%**; LEV-7: **8.24×** (synthetic, not a production baseline).
+
+These numbers are self-assessment plus the repository's evidence chain, not a third-party review — the report's `notClaimed` field lists exactly what is not claimed: official standard, certification, security audit, ranking, production baseline.
+
+<details>
+<summary>Repository map (the numbered 00–99 specs are written in Chinese)</summary>
+
+- [START-HERE.md](START-HERE.md) — the one-page entry for a new session
+- [docs/DEVELOPER-JOURNEY.md](docs/DEVELOPER-JOURNEY.md) — the plugin lifecycle and toolchain
+- [spec/levels.json](spec/levels.json) — the machine source of truth for the three levels
+- [scheme/SCHEME-v1.md](scheme/SCHEME-v1.md) — the A–E / A–D evaluation scheme and rubric
+- [evidence/README.md](evidence/README.md) — the append-only evidence chain and how to re-run it
+- [GOVERNANCE.md](GOVERNANCE.md) · [CONTRIBUTING.md](CONTRIBUTING.md) · [SECURITY.md](SECURITY.md) · [MODIFICATION-POLICY.md](MODIFICATION-POLICY.md)
+- [RENAMES.md](RENAMES.md) — old-name to new-repository map
+- [P0-P2-STATUS.md](P0-P2-STATUS.md) — item-by-item DONE / PARTIAL / REPLACED
+
+The full 51-document index lives in the Chinese [README.md](README.md).
+
+</details>
+
+## Three-step setup
+
+**1. Install** (either way; Node `>=24.9.0`):
+
+```sh
+# global install: both dsh-standards and dsh-levels land on PATH
+npm i -g dsh-standards
+# or run without installing anything
+npx dsh-standards --version
+npx -p dsh-standards dsh-levels --help
+```
+
+There is no build step: `bin/` and `40-checker/` are the shipped artifacts. The second npx command must use `-p dsh-standards`, because the bare npm names `dsh-levels` and `dsh-plugin-verify` belong to other projects.
+
+**2. Run one check** (the target is a package directory containing `package.json`):
+
+```sh
+# entry compatibility: static recognition, no runtime claim
+dsh-standards ./my-plugin --scope entry --entry-claim static
+
+# full DS-C (default scope): JSON / SARIF 2.1.0 reports
+dsh-standards ./my-plugin --format json --out report.json
+dsh-standards ./my-plugin --format sarif --out report.sarif
+
+# three-level audit: ENTRY runs automatically; FLOW / BOOST stay not-run / not-measured without evidence files
+dsh-levels ./my-plugin --format json --out audit.json
+```
+
+**3. Read the verdict** — exit code first, then `status`:
+
+```text
+0  completed, no FAIL (PASS or WARN; WARN does not claim a clean PASS)
+1  completed, at least one FAIL
+2  incomplete: a mandatory check did not run, or the checker itself failed
+3  usage error
+```
+
+Real output on this repository's fixture (trimmed):
+
+```text
+scope=entry profile=package status=pass exit=0
+level: DSH-ENTRY recognized
+DSC-01  PASS  C-1    L0
+DSC-14  NA    C-14   L1     entry static claim: no runtime evidence
+coverage: mandatory=5 pass=5 warn=0 fail=0 na=0 notRun=0
+```
+
+On the same fixture `dsh-levels` reports the three layer states — `entry.state=recognized`, `flow.state=not-run`, `boost.state=not-measured` — plus three `nextActions` lines telling you which evidence to supply next.
+
+## Core capabilities
+
+- **Static face (L0)**: never executes the target; reads the manifest, entry, patch, and sources, and decides `DSC-01`–`DSC-03` and `DSC-10`–`DSC-13`.
+- **Runtime face (L1–L3)**: loads the package into a Cordis runtime inside an **isolated child process** and checks that it reaches ACTIVE, that the declared `provide` is actually registered, that effects form a topology, that unload returns to zero, and that restart is idempotent. A timeout is a hard kill, and a kill is `not-run`, never PASS.
+- **Report face**: `text` / `json` / `sarif` (SARIF 2.1.0, ready for GitHub code scanning); each report carries an `anchor` (harness commit, dsh, cordis versions) — change the version and you are holding a different ruler.
+- **Three-level audit**: `dsh-levels` maps the ENTRY checker verdict onto an evidence state and only reads J1–J8 (FLOW) and LEV-1–8 (BOOST) evidence files; without evidence it reports `not-run` / `not-measured`.
+- **Read-only**: it does not modify the inspected package and writes nothing into the target directory; reports go to stdout or to `--out`.
+
+### Checks
+
+| ID | Clause | Layer | What it checks |
+|---|---|---|---|
+| `DSC-01` | C-1 | L0 | `manifest`, `name`, `version`, `license`, entry, `dsh.bundle.patch`, export shape |
+| `DSC-02` | C-2 | L0 | `engines.dsh` or the `@deepseek-ai/dsh*` peer, satisfied by the runtime |
+| `DSC-03` | C-3 | L0 | install spec, market index entry, supply-chain pinning |
+| `DSC-04` | C-4 | L1 | loads to ACTIVE |
+| `DSC-05` | C-5 | L1 | declared `provide` really registers |
+| `DSC-06` | C-6 | L1+L2 | effect topology, private registration surface |
+| `DSC-07` | C-7 | L2 | effects / stores / registries return to zero after unload |
+| `DSC-08` | C-8 | L3 | restart idempotence plus cold restart |
+| `DSC-09` | C-9 | L1 | declared vs measured cross-check |
+| `DSC-10` | C-10 | L0 | import-time side effects (advisory, never FAIL) |
+| `DSC-11` | C-11 | L0 | `dsh.capabilities` plus high-signal triggers |
+| `DSC-12` | C-12 | L0 | credential backing file, `listRecords`, global seams; also reads the shape of `dsh.commitments` and `dsh.disclosure` |
+| `DSC-13` | C-13 | L0 | `dsh.labels`, `upstream`, forbidden wording |
+
+`--scope mvp` only covers 8 IDs (`DSC-03/10/11/12/13` are outside the MVP) and writes `subset: "MVP-13"` with `level: null` — that is **not** a DS-C pass.
+
+### Which tool, when
+
+| Stage | Command | Tool | Output |
+|---|---|---|---|
+| Start | `dsh-dev new` / `dsh-start new` | dsh-plugin-start | installable skeleton |
+| Static precheck | `dsh-dev check` / `dsh-ready` | dsh-plugin-ready | static report (host `not-run`) |
+| Host contract | `dsh-clinic --only R,K,X` | dsh-plugin-doctor/clinic | host contract |
+| Isolated install | `dsh-test-drive` | dsh-test-drive | C3 evidence |
+| Market submission | `entry-validate --host-evidence` | this repo's `40-checker` | C1–C8 + C3 |
+| Tool management | `dsh-tools list/doctor/install` | dsh-plugin-tools | versions / install guidance |
+
+The full journey is in [docs/DEVELOPER-JOURNEY.md](docs/DEVELOPER-JOURNEY.md), the tool family in [dsh-plugin-devtools](https://github.com/PerryLink/dsh-plugin-devtools), and the old-name map in [RENAMES.md](RENAMES.md).
+
+## Configuration
+
+There is no config file — every knob is a command-line flag. The two commands have separate sets: `dsh-standards <package-dir>` forwards to the DS-C checker (`--help` prints the full usage), and `dsh-levels <package-dir>` forwards to the three-level audit.
+
+The eight you will actually use:
+
+| Flag | Default | Meaning |
+|---|---|---|
+| `<package-dir>` | `.` | target package directory containing `package.json`; tarballs are not supported |
+| `--scope` | `ds-c` | `mvp` = 8 MVP IDs; `ds-c` = full claim; `entry` = compatibility gate |
+| `--format` | `text` | `text` / `json` / `sarif` (SARIF 2.1.0) |
+| `--entry-claim` | `runtime` | `runtime` needs DSC-14 runtime evidence; `static` claims structural recognition only |
+| `--host` | — | app-boot `lib/index.js`; required for the bundle G0–G3 smoke |
+| `--profile` | `package` | `package` / `listing` / `both`; defaults to `listing` when `--market-index` is given |
+| `--no-exec` | `false` | skip L1–L3 entirely, reported as `not-run` (never PASS) |
+| `--out` | — | write the report to a file instead of stdout |
+
+<details>
+<summary>Full flag table (every dsh-standards switch)</summary>
+
+| Flag | Default | Meaning |
+|---|---|---|
+| `--format` | `text` | report format: `text` / `json` / `sarif` |
+| `--scope` | `ds-c` | check scope: `mvp` / `ds-c` / `entry` |
+| `--profile` | `package` | evaluation profile: `package` / `listing` / `both` |
+| `--runtime` | — | path to a built `@deepseek-ai/cordis` `lib/index.js`; enables L1–L3 |
+| `--runtime-version` | `0.2.2-alpha.2` | dsh runtime version recorded in the anchor and used by C-2 |
+| `--cordis-version` | `4.0.6-alpha.1` | cordis version recorded in the anchor |
+| `--commit` | `ce7073d51254` | harness commit recorded in the anchor |
+| `--install-spec` | — | recommended install spec for C-3 |
+| `--market-index` | — | format-1 market index JSON; turns on listing mode |
+| `--plugin-name` | — | plugin name inside the market index |
+| `--timeout` | `10000` | L1–L3 timeout in ms (100–120000) |
+| `--no-exec` | `false` | skip L1–L3 entirely, reported as `not-run` / `NA` |
+| `--static-only` | `false` | M1 mode: only C-1/C-2/C-3/C-11/C-12/C-13 stay in scope, C-4–C-9 are out |
+| `--treat-as-plugin` | `false` | run L1–L3 against a bundle's server entry |
+| `--entry-runtime` | `require` | entry gate: full G0–G3, or accept the `server-entry` proxy |
+| `--entry-claim` | `runtime` | entry gate: runtime evidence (DSC-14), or static recognition only |
+| `--host` | — | app-boot `lib/index.js`; enables the bundle G0–G3 smoke |
+| `--base` | — | base `cordis.patch.yml` for the bundle smoke |
+| `--web` | — | web-app `cordis.patch.yml` for the bundle smoke |
+| `--bare` | — | bare-module base directory for the host farm |
+| `--out` | — | write the report to a file instead of stdout |
+| `--help` / `--version` | — | print usage / print `dsh-standards <version> (checker 0.1.0)` |
+
+`dsh-levels` flags:
+
+| Flag | Default | Meaning |
+|---|---|---|
+| `<package-dir>` | required | target package directory containing `package.json` |
+| `--host` | — | real G0–G3 boot for ENTRY |
+| `--base` | — | base patch for the bundle smoke |
+| `--bare` | — | bare-module farm for the bundle smoke |
+| `--no-runtime` | `false` | stay static for ENTRY even when `--host` is given |
+| `--flow-evidence` | — | J1–J8 evidence JSON; each entry has `status` set to `pass` / `fail` / `not-run` |
+| `--boost-evidence` | — | LEV evidence JSON; a row with a status but no evidence card is downgraded to `not-measured` |
+| `--timeout` | `150000` | checker timeout in ms |
+| `--format` | `text` | `text` / `json` |
+| `--out` | — | write the result to a file |
+
+</details>
+
+## FAQ
+
+<details>
+<summary>Is this a certification or a ranking?</summary>
+
+No. The project describes itself as **non-official, not a certification, not a security audit**, and it does not represent an endorsement by DeepSeek; it charges no money and sells no levels or rankings. `A–D` are quality grades from `scheme/`, orthogonal to the `ENTRY` / `FLOW` / `BOOST` levels — neither reports nor docs merge them into one score. However strong the evidence, it is self-assessment plus an in-repo evidence chain, and the `notClaimed` field always lists "official standard / certification / security audit / ranking / production baseline".
+
+</details>
+
+<details>
+<summary>What is the difference between `18/18 recognized` and `18/18 clean`?</summary>
+
+`recognized` is a static recognition pass (`--entry-claim static`, structural claim only, no runtime claim). `clean` needs runtime evidence: base-only 10/18 carry runtime evidence (8 clean + 2 reject), the real web profile has 16/18 warning-free with 2 host-gate rejects, and all of that is self-assessed without independent review — so **18/18 clean is still not achieved**, and no document claims otherwise. Likewise, `client-module-loaded` means neither UI usability nor certification.
+
+</details>
+
+<details>
+<summary>What happens without a DSH runtime installed? Does it pass silently?</summary>
+
+It does not. L1–L3 need `--runtime` (and a bundle needs `--host`); without them the checks are reported as `ERR` / `not-run` with the reason inline (`no --runtime supplied; L1-L3 not run`), and under the package scope the conclusion becomes `incomplete` with exit `2`. `--scope entry --entry-claim static` still yields `DSH-ENTRY recognized`, but that is a static claim; pass `--runtime` or `--host` for a runtime verdict.
+
+</details>
+
+<details>
+<summary>Does it touch my package? Can it break my code?</summary>
+
+The checker only reads the target directory and never modifies it. When runtime checks run, loading happens in a **child process** with a hard kill on timeout (10s by default, adjustable with `--timeout`); a kill is recorded as `not-run`, not PASS, and the child cannot read the checker's own environment. Add `--no-exec` to skip execution entirely, or `--static-only` for a static-only verdict.
+
+</details>
+
+<details>
+<summary>How does this relate to guide / doctor / cert / portal / catalog?</summary>
+
+This repository is the specification text plus the checker and the level audit; the prose guide lives in `dsh-plugin-guide`, host-contract measurement lives in `dsh-plugin-doctor` (clinic), and the market and portal tools are separate. The role split and contribution paths are in the Chinese `39` document, and the selection table above summarises when to reach for which tool.
+
+</details>
+
+## Known limitations
+
+- **No certification and no security audit.** A report is a quality signal, not a compliance proof; `A–D` are quality grades, not `ENTRY` / `FLOW` / `BOOST`.
+- **No fees, no level sales, no ranking sales.**
+- **Never writes static up as runtime.**
+- **Never writes 10/18 base-only up as 18/18 web/production, and never writes `client-module-loaded` up as UI or certification.**
+- **Never claims official recommendation or adoption, certification, accreditation, or a security audit.**
+- **MVP coverage gap**: `--scope mvp` covers only 8 IDs (`DSC-03/10/11/12/13` are outside it); its verdict is `subset: "MVP-13"` with `level: null` and must not be used as a DS-C badge.
+- **Package directories only**: the target must be a directory containing `package.json`; tarball support is not part of the MVP.
+- **Runtime verdicts have prerequisites**: without `--runtime` or `--host`, L1–L3 can only be `not-run`, and `--entry-claim static` can never produce a runtime verdict.
+- **Verdicts are anchored to versions**: change `commit` / `dsh` / `cordis` and you are holding a different ruler; an old report is not evidence for a new version.
+- **BOOST numbers are not a production baseline**: the 8/8 measured LEV results come from loader / context / provider-capture / synthetic-timer environments, and the LEV-7 8.24× figure likewise.
+- **Text changes must be declared**: copying or modifying text from this repository requires a prominent Change Notice (source commit, which clause, what changed, why, date), see [`MODIFICATION-POLICY.md`](MODIFICATION-POLICY.md); marks and badges are not licensed by the licenses above, see `governance/mark-policy.md`.
+
+## Development
+
+```sh
+git clone https://github.com/PerryLink/dsh-standards.git
+cd dsh-standards
+
+node --test 40-checker/test/*.test.mjs                      # 128/128
+node bin/dsh-standards.mjs 40-checker/fixtures/good \
+  --scope entry --entry-claim static --format text          # fixture smoke
+node bin/dsh-levels.mjs 40-checker/fixtures/good --format json
+
+# straight to the checker from a source checkout
+node 40-checker/bin/dsh-plugin-verify.mjs <plugin-dir> --scope entry --entry-claim static
+node 40-checker/bin/dsh-plugin-verify.mjs <bundle-dir> --scope entry \
+  --host /path/to/@deepseek-ai/dsh-app-boot/lib/index.js
+
+node 40-checker/scripts/journey-runtime.mjs --package <dir> --json   # J1 install/uninstall journey
+node 40-checker/scripts/governance-check.mjs                # governance clauses
+node 40-checker/scripts/attestation-check.mjs               # attestation records
+node 40-checker/scripts/claim-check.mjs                     # claim consistency
+node 40-checker/scripts/dco-check.mjs --range HEAD~1..HEAD  # DCO
+node publish/verify.mjs --id DSH-ATT-2026-001               # verify an example record
+
+cd examples/boost && node lev7-bench.mjs 64 20              # LEV-7 bounded-parallelism benchmark
+```
+
+README images are generated programmatically by [`scripts/gen-readme-assets.py`](scripts/gen-readme-assets.py) (Python + Pillow, light and dark SVGs) and can be regenerated any time: `python scripts/gen-readme-assets.py`.
+
+### Layout
+
+```text
+00-99*.md         specification / research / profile / communication docs
+bin/              two public entry points: dsh-standards, dsh-levels
+40-checker/       checker, J1 journeys, bundle smoke, samples, results
+governance/       roles, contribution terms, mark policy, appeals, COI, decision log
+adoption/         official / market / developer adoption packs
+spec/             levels.json + clauses.json + schemas (evidence / report / attestation)
+scheme/           A–E / A–D evaluation scheme and rubric
+evidence/         append-only evidence chain and re-run instructions
+examples/boost/   LEV-7 reference implementation and benchmarks
+publish/          label examples + attestation record JSON
+docs/             long-lived docs such as DEVELOPER-JOURNEY plus the README images
+```
+
+### Governance and contribution
+
+- Roles and permissions: `governance/roles.json`; one-page summary: [GOVERNANCE.md](GOVERNANCE.md)
+- Contribution: DCO 1.1 + DCT v1 (`dsh-ct/1`, a licence rather than a copyright assignment), see [CONTRIBUTING.md](CONTRIBUTING.md)
+- Marks: `governance/mark-policy.md` (auto-granted, revocable, disclaimer required)
+- Appeals: `governance/appeals.md`; AI only receives, times, and reviews — humans decide
+- Three adoption packs: `adoption/{official,market,developer}/`
+
+## Family and credits
+
+This project is one member of the DSH ecosystem maintained by [PerryLink](https://github.com/PerryLink): this branch owns the ruler — the specification text, the checker, and the level audit — while the guide, scaffolding, verification, and portal tools have their own repositories. If it saved you a few manual cross-checks, **give it a star** — it unlocks nothing, but it helps the next person who has to verify a plugin find this repository.
+
+The earlier project `dsh-plugin-standard` (lanbaolu) belongs to the same ecosystem and is not affiliated with this project; any reference or comparison will cite its source.
+
+## License
+
+- **Text / specifications / documentation / translations**: CC BY 4.0 ([`LICENSE-CC-BY`](LICENSE-CC-BY)); modifications must carry a Change Notice per [`MODIFICATION-POLICY.md`](MODIFICATION-POLICY.md);
+- **Code / scripts / schemas**: Apache-2.0 ([`LICENSE-APACHE`](LICENSE-APACHE), including the patent grant and defensive termination);
+- **Marks / badges**: separate rules in `governance/mark-policy.md`, not granted by the licenses above;
+- **Contributions**: DCO 1.1 + DCT v1, with no copyright assignment.
+
+<p align="center"><sub>CC BY 4.0 (text) · Apache-2.0 (code) · community project, not affiliated with DeepSeek.</sub></p>