last30days-skill-cn
Jesseovo
last30days-cn 是一个 AI Agent 技能(Skill),能够自动搜索中国互联网 8 大主流平台最近 30 天的内容,综合分析后生成有据可查的研究报告。
PROJECT TOPICS
INSTALL REFERENCE
dsh plugin --profile web add github:qiuyiwu1989-star/dsh-k12-substrate
该命令指向仓库当前默认分支;尚无绑定当前 commit 的完整验证结果。
PROJECT README
English | 中文
A K12 capability substrate for DeepSeek Harness. It lets the model look up capability anchors from China's national curriculum standards, check whether a character/word/poem is on an official list, record what a specific child has mastered (locally), and compute their character count and next steps.
Data comes from China's MOE Compulsory Education Curriculum Standards (2022).
This section comes first, because mistaking this for teacher-validated data is worse than not installing it.
This package ships 4,231 usable anchors across 24 subjects (source: data/substrate.json, built from substrate v1.12). What "usable" means matters more than the number:
| count | what it means | |
|---|---|---|
| Objectively decidable | 223 | correctness is mechanical — a character is written correctly or it isn't |
| Judged by AI against the verbatim standard text | 4,008 | an AI read the source page and found no fault |
| ↳ of which our own capability rewrites | 912 | the standard said "know / understand X"; we wrote a checkable ability on top of it. This is our educational judgment, not the standard's wording — it traces back to a source sentence, not to a verbatim requirement |
| Signed off by a teacher | 0 | none |
So "usable" means "an AI checked it and found no fault" — not "a teacher approved it." Treat every anchor that way.
Math, physics, chemistry and the rest are now included (math: 412). There are 4,823 dependency edges among usable anchors, so the plugin can answer "what should be learned before this" — but those edges were proposed by a model and retagged in two stages, not validated by a teacher either.
One more boundary: anchors are at the granularity of the national standard, not of a lesson. Most span a whole stage (e.g. grades 1–2). Mapping a single lesson to one tells you where it sits, not that the child has finished learning it. Each result carries a granularity warning for this reason.
pnpm add dsh-k12-substrate
pnpm dsh web --patch ./node_modules/dsh-k12-substrate/cordis.patch.yml
Or insert into your own cordis.yml:
- insert:
- id: k12-substrate
name: 'dsh-k12-substrate'
config:
profileDir: '' # empty → ~/.dsh-k12-substrate/profiles/
readOnly: false # true → register query tools only, no profile writes
| Tool | What it does |
|---|---|
k12_substrate_info |
Coverage, provenance, and known limitations. Call before asserting what the standards require |
k12_find_capability |
Search anchors by subject / grade band / keyword; returns decidable statements and the basis for each |
k12_lookup_item |
Locate a character, word, or recitation piece in the official appendix lists: which table, what index, which grade band, recognize vs. write |
k12_record_mastery |
Record what a learner has mastered, to a local profile file |
k12_learner_progress |
Character count, vocabulary size, pieces recited, per-anchor completion, and next items |
The last two can be disabled with readOnly: true.
When the host mounts the substrate's MCP server (via the official bridge @deepseek-ai/dsh-mcp-client, serverName: k12), k12_find_capability with a query searches through it: live substrate data, the substrate's own ranking, a grain warning on every hit, and the hit is recorded by the substrate's usage counter (anchor IDs and counts only). Without it, the tool falls back to the bundled snapshot. Which path was taken is stated in the route field and on the first line of the result — never a silent fallback.
K12_TAXONOMY_ROOT=/path/to/os-k12-taxonomy \
dsh web --patch ./examples/mcp/cordis.patch.yml --patch ./cordis.patch.yml
What goes through MCP, what doesn't, and the counting rules: docs/mcp.md (Chinese).
1. Profiles stay local, and only counts are echoed back.
k12_record_mastery records what a specific child can and cannot do — a minor's learning profile. It writes only under profileDir (default ~/.dsh-k12-substrate/profiles/) and uploads nothing. The tool result reports counts, not the individual items: tool results enter the model's context and may be written to session logs, compacted, and sent to the model provider. Counts are what a product needs; item-by-item detail is not.
learner becomes the filename, so use a pseudonymous id (e.g. stu_0001). Values containing path-traversal characters are rejected.
2. Model judgements are always proposed.
Only a holder starting with teacher: or parent: is recorded as confirmed. The model saying a child knows something does not make it so, and silence is not confirmation. k12_learner_progress reports the two separately.
3. Only usable anchors can be referenced. An assertion pointing at an unreviewed anchor means measuring a child with an unvalidated ruler. Passing an ID outside the usable set raises an error.
pnpm install
pnpm snapshot # rebuild the data snapshot from ../os-k12-taxonomy
pnpm verify # typecheck + smoke assertions (no DSH runtime, never starts the real MCP server)
pnpm build
The smoke test (scripts/smoke.ts) runs without the DSH runtime, calling execute() directly against real data, real boundaries, and real disk writes. It caught two things typecheck could not: required on a node referenced by items throws UNSUPPORTED_SCHEMA at runtime; and a dedupe key format duplicated across three files where one separator was a literal NUL byte.
Code MIT; bundled data ODbL v1.0 + CC BY-SA 4.0 (matching upstream). The curriculum text itself is not redistributed and remains the property of China's Ministry of Education. See LICENSE.
CLASSIFICATION EVIDENCE
系统优先读取 GitHub Topics,再与站内分类词典和词根规则比对。当前命中: education、knowledge-graph。