@aifeed/deepseek-harness
DeepSeek Harness 插件，提供 AIFeed 工具：验证签名内容权限、获取受令牌预算限制的 Markdown、选择增量索引条目，并决定是否使用。
插件会安装到这里；不确定时保持 web。
npx -y @deepseek-ai/dsh plugin --profile web add @aifeed/deepseek-harness@1.0.0-draft.2Deepseek Harness 以 @aifeed/deepseek-harness 发布，当前版本为 1.0.0-draft.2。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。
English · Bahasa Indonesia · 中文
Cordis plugin that gives the DeepSeek Harness six model-callable tools for working with
AIFeed: signed content permissions for the AI web. It reuses the same
zero-dependency engine as aifeed-mcp-server — the
engine/ copies (server.js, lib, schema) are generated with npm run build:harness from
the repository root.
| Tool | What it does |
|---|---|
aifeed_verify_manifest | Verify a manifest: signature, DNS _aifeed anchor, result (VERIFIED/UNVERIFIED) |
aifeed_fetch_aifeed | Fetch a page as token-budgeted AIFeed Markdown/MAKO with permissions + optional signature check |
aifeed_list_assets | List the images, videos, and documents a signed page declares |
aifeed_verify_asset | Download a declared asset and check size/sha-256 |
aifeed_select_index | Rank signed delta-index entries within page/token budgets |
aifeed_decide_usage | Decide whether a usage (retrieval, training, …) is allowed |
dsh plugin --profile web add @aifeed/deepseek-harness
dsh --profile web --dump-config # shows the bundle layer
dsh web
The package ships a bundle layer (cordis.patch.yml), so the tools activate as soon as the
profile lists it. To try a checkout instead:
dsh plugin --profile demo add ./integrations/deepseek-harness
Then ask the agent, for example:
Use aifeed_decide_usage for shop.aifeed.md with usage "training".
| Option | Default | Description |
|---|---|---|
allowPrivate | false | Permit http:// loopback origins for local testing (sets AIFEED_MCP_ALLOW_PRIVATE=1 in the harness process) |
Only https:// origins are fetched by default. Cancellation follows the harness
exec.signal at the call boundary; in-flight fetches finish on their own timeout.
The engine copies (engine/server.js, engine/lib/, engine/schema/, LICENSE) are generated from
packages/aifeed-mcp-server/:
npm run build:harness # regenerate
npm run harness:check # verify in sync