mirror of
https://github.com/deepseek-ai/deepseek-harness
synced 2026-08-15 21:04:50 +00:00
5f6ccf91ec
Codex adversarial review flagged that "guidance not authority" over-rotated: it could be read as licensing a reviewer to waive genuinely-blocking checks (HMR-safety test, doc-sync, ADR 0007 gates), and the "Where to look first" section claimed to list things "not in the docs" while its first two items ARE documented requirements. Fix both: scope independent judgment to *what to look at* and *how to apply a rule*, not whether documented requirements hold; state AGENTS.md / packages/AGENTS.md / ADR 0007 stay authoritative. Split the checks into "Hard blockers (documented requirements)" and "Reviewer-only checks (judgment required)", so doc-sync/HMR-safety/gates aren't mixed with reviewer-only heuristics. Keep the independent/broad-reasoning intent.
DeepSeek Harness
Monorepo for the DeepSeek Harness group.
Projects
- DeepSeek Code — DeepSeek's coding agent product.
Development
This monorepo is built on the Cordis framework (vendored as source under vendor/), microkernel-style: everything is a plugin.
pnpm install
pnpm run test # vitest
pnpm run demo:echo # runnable echo-agent example (no API key needed)
pnpm run demo:coding # the real DeepSeek coding agent (needs DEEPSEEK_API_KEY)
For humans, start with the development guide for local setup, hooks, environment variables, and quality gates, then read the architecture design before package work. Local context lives in packages/ and vendor/.
For agents, follow AGENTS.md.