Skip to content

feat(agents): add financial-engineer (20th subagent) + drain #311–#330 doc drift#331

Merged
dackclup merged 2 commits into
mainfrom
claude/sharp-turing-q3A5J
May 31, 2026
Merged

feat(agents): add financial-engineer (20th subagent) + drain #311–#330 doc drift#331
dackclup merged 2 commits into
mainfrom
claude/sharp-turing-q3A5J

Conversation

@dackclup
Copy link
Copy Markdown
Owner

Summary

Two coupled deliverables (they share the same six current-state doc homes, so splitting would force a rebase on each merge — the §Phase-status-collision lesson). Doc + agent-infra only — no compute / schema / scoring / valuation / frontend code change.

(a) New 20th subagent — financial-engineer

The team's first generative quant-design seat — the design counterpart to methodology-scientist's validation seat. The roster had reviewers / validators / auditors but no seat that proposes a new quant construct.

  • Tier 3 Specialized · opus · read-only (Read/Bash/Grep/Glob — no Edit/Write, mirrors methodology-scientist)
  • Emits a structured design proposal (problem → academic anchor → math spec → architecture fit → annotate-before-veto rollout → observability fields → test plan → orthogonality check → footguns) for a new valuation method / factor signal / scoring pillar / risk-overlay defense flag / cost-of-equity refinement
  • Charter Phase 4–7 — factor consolidation · ML meta-learner · sentiment v2 · regime + portfolio construction — inside the fundamentals-equity-ranking research framework (no TA / options-flow / HFT)
  • Handoff chain: financial-engineer (design) → methodology-scientist (ratify the prior, can REJECT) → test-engineer (tests) → quantrank-reviewer (review). New README Flow 8 (quant-design) is the generative complement to Flow 3 (new-defense flow). Designer proposes; validator gates — separate seats so no agent both invents and ratifies its own prior.
  • Carries the project discipline in-prompt: academic-anchor-or-labeled-gut-feel · annotate-before-veto · observability-before-wiring · scout-then-integrate · orthogonality (φ) · sum-to-1 · graceful-degradation.

User-locked this session (AskUserQuestion): model = opus (5th opus agent; generative quant design is breadth-of-judgment, rare-fire so small all-models-pool impact) · charter = broad Phase 4–7.

(b) Doc-drain housekeeping

Session-start orientation found the docs trailing the repo: PHASE_STATUS.md §Recently merged stopped at #310 (8 PRs behind HEAD = #330) and CLAUDE.md stopped at #326.

Verification

Gate Result
ruff check . (whole repo) ✅ All checks passed
pytest -m "not network" N/A — zero Python touched (trivially green)
schema_check N/A — no schema field moved
tsc --noEmit / next build N/A — no frontend code touched
Subagent-count lockstep grep ✅ no current-state 19/4 opus/7 flows left
New agent vs README §Authoring conventions ✅ all 6 (one-job · sharp TRIGGER+skip-guards · opus-justified · read-only allowlist · project-anchored · pinned output + Handoff)

What this PR does NOT touch

  • No compute/ (ingest / scoring / valuation / output / main)
  • No frontend/ code, no schemas.py / types.ts / schema-snapshot.json
  • No .github/workflows/, no deps
  • Defense layer unchanged (33 declared flags, 7 vetoes); schema stays 0.10.11-phase4.6

Reviewers (in flight)

Spawned read-only at the pre-push gate: docs-reviewer (cross-file substance/consistency + new-agent conventions) + phase-coordinator Mode B (CLAUDE.md+AGENTS.md lockstep + branch-collision). Any findings will land as follow-up commits on this draft.

https://claude.ai/code/session_0148EoMmL6zakDWqHXjqQ9yq


Generated by Claude Code

…doc drift

Adds the team's first GENERATIVE quant-design subagent — the design
counterpart to methodology-scientist's validation seat. The roster had
reviewers / validators / auditors but no seat that PROPOSES a new quant
construct. financial-engineer (Tier 3 Specialized, opus, read-only) emits
a structured design proposal (problem -> academic anchor -> math spec ->
architecture fit -> annotate-before-veto rollout -> observability fields
-> test plan -> orthogonality -> footguns) for a new valuation method /
factor / scoring pillar / defense flag / cost-of-equity refinement, with a
charter spanning Phase 4-7 (factors / ML / sentiment / regime + portfolio).
It hands off down the chain methodology-scientist (ratify the prior) ->
test-engineer (tests) -> quantrank-reviewer (review). New README Flow 8
(quant-design) is the generative complement to Flow 3 (new-defense flow).

User-locked this session: model = opus (5th opus agent; generative quant
design is breadth-of-judgment, rare-fire so small all-models-pool impact);
charter = broad Phase 4-7.

Also drains the doc drift session-start orientation surfaced: PHASE_STATUS.md
Recently-merged was 8 PRs behind (stopped at #310; HEAD #330) and CLAUDE.md
stopped at #326. Drained #311-#330 into PHASE_STATUS.md + #327-#330 into
CLAUDE.md. Subagent counts updated in lockstep across all six current-state
doc homes: 19->20 agents, 4->5 opus, 7->8 flows, Tier 3 5->6.

Doc + agent-infra only — no compute / schema / scoring / valuation /
frontend code change. ruff clean; no Python touched.

https://claude.ai/code/session_0148EoMmL6zakDWqHXjqQ9yq
@vercel
Copy link
Copy Markdown

vercel Bot commented May 31, 2026

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
quantrank Ready Ready Preview, Comment May 31, 2026 5:49am

…awn-freq list

docs-reviewer FAIL #1: the §spawn-frequency paragraph's opus-agent list
still named only 4 (`incident-commander` · `release-captain` ·
`methodology-scientist` · `quantrank-reviewer`) — a stale reference the
20th-subagent lockstep pass missed (it doesn't carry the literal "four
opus" prefix the count grep matched on). Added `financial-engineer` as
the 5th. This was the only current-state opus-quad list left stale; the
CLAUDE.md §Spawn-discipline copy was already fixed, and the CLAUDE.md
§Phase-status "4 opus" mentions are historical PR-notes (left as-is).

Also resolves docs-reviewer WARN #2: tightened "~3.5k lines across all of
`.claude/agents/`" → "across the 20 agent files" so the figure (3528 lines,
the 20 prompt files, excluding README) is unambiguous and accurate.

https://claude.ai/code/session_0148EoMmL6zakDWqHXjqQ9yq
@dackclup dackclup marked this pull request as ready for review May 31, 2026 05:54
@dackclup dackclup merged commit 3cb95eb into main May 31, 2026
4 checks passed
@dackclup dackclup deleted the claude/sharp-turing-q3A5J branch May 31, 2026 05:54
dackclup pushed a commit that referenced this pull request Jun 2, 2026
…In-flight entries

- CLAUDE.md §Phase status: new "Recently merged" block (PR #331#373,
  2026-05-31→2026-06-02, 43 PRs covering $impeccable full-frontend pass +
  financial-engineer subagent + PR-B country/exchange chips + Commit A
  MUST-FIX sweep); old #303#330 block demoted to new "Earlier" section;
  "In flight" entries (PR-B + financial-engineer) replaced with "(none)"
- PHASE_STATUS.md: date 2026-05-31 → 2026-06-02; same recently-merged /
  earlier restructure with compact one-liners for all 43 PRs
- AGENTS.md §Phase + version state: "In flight — PR-B" bullet replaced
  with "all PRs through #373 merged" pointer to CLAUDE.md
- PHASE_STATUS_INFLIGHT.md: Commit B entry appended

No compute / schema / scoring / valuation / frontend code change.

https://claude.ai/code/session_01ELWfJoJp5kMje2j4zoUCQh
dackclup added a commit that referenced this pull request Jun 2, 2026
…ntries (#380)

Drains 43 PRs (#331#373) into CLAUDE.md + PHASE_STATUS.md Recently-merged blocks. Old #303#330 block demoted to Earlier. In-flight entries for PR-B and financial-engineer replaced with (none). AGENTS.md In-flight bullet updated. PHASE_STATUS_INFLIGHT.md Commit B entry appended. No compute/schema/frontend change.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants