Skip to content

dsh-rigorquant v0.1.0

Choose a tag to compare

View all tags
@linxichen linxichen released this 14 Aug 20:30
· 11 commits to master since this release
v0.1.0
8c7a810

RigorQuant v0.1.0

Initial release of the RigorQuant agent preset + bundled skill for DeepSeek Harness: an unattended, walled multi-agent research framework for empirical/computational mathematics (econ, finance, portfolio, simulation).

What's included

  • Walled multi-agent protocol — parallel explorers (blank-context subagent), a dual-derivation ground-truth track, and a counterexample-only adversary.
  • Four-part check battery — closed-form equality, exact invariants, analytic bounds, and staged statistical hardening, run on simplified cases before numerical implementation.
  • jacobian MCP escalation lane (with Lean as the last resort) for proof-critical claims.
  • Pinned uv compute lane (sympy/mpmath/cvxpy/hypothesis/jax) with an exact-version lockfile.
  • PASS / BLOCKED / BUDGET lifecycle with terminal honesty.
  • Study workspace (Mode A/B) — one study per repo or studies/<slug>/, durable deliverables committed, scratch in a gitignored interim/.
  • dsh.bundle manifest for dsh plugin add ecosystem installs.

Install

# bundle (skill layer)
dsh plugin --profile web add github:linxichen/dsh-rigorquant

# or full preset
git clone https://github.com/linxichen/dsh-rigorquant
cd dsh-rigorquant && ./install.sh

See README.md and docs/architecture.md for the full design record.

Assets 2
Loading