9 Sub-Agent Chain · Routes on /fpr TICKER
#1
Data Ingest — raw financials + source log
raw_financials.json
#2
Reconstruction + Adjustments — Reported / Adjusted / Normalised
adjusted_financials.json
#3
Returns + DuPont + Balance Sheet
returns_analysis.json
#3B
WACC build — CAPM + capital structure
wacc.json
#3C
Sanity Check Gate — arithmetic + reconciliation
HALT if fails
#4
Sector Adapter + Peer Benchmark — 8 adapters
peer_benchmark.json
#5
Forecast Gap + Variant View — alpha signal vs consensus
variant_view.md
#6
Scoring + Red Flags + Report
scoreboard.json · report.docx
#7
Change Detection — delta vs prior run
cold-start tags
Standing test ticker — TDOC (Teladoc Health Inc)
On Tier 4 data (WebSearch), expect INSUFFICIENT_DATA halt at ~52% weighted coverage. That is correct behaviour. Rule #13 EQ cap applies (SBC >15%). Sector routes
tech_saas → healthtech (BETA). Full baseline in reference_runs/TDOC_v1.3.1_validation.md.
Your ask — 20 minutes, tonight
Drop → /fpr TDOC → grade 5 gates → report back
- Copy
fpr-module/intopineal-capital/.claude/skills/ - Read
SKILL.md+LIMITS.md(5 min) - Run
/fpr TDOCin Claude Code - Grade output against 5 validation gates below
- Reply with pass/fail + any methodology flags for v1.3.2
5 Validation Gates
| G | Gate | Pass criterion |
|---|---|---|
| G1 | Coverage | Rule #12 fires at <60% weighted; does not silently proceed |
| G2 | Adapter | Sector adapter declared up-front with route reasoning |
| G3 | EQ cap | Rule #13 fires (SBC >15%); total cap ≤60 applied |
| G4 | Halts | Halt taxonomy used; no orphan flags, no silent failures |
| G5 | No fabrication | Every field sourced or explicit N/A; no "approximately" |
Expected on TDOC tonight: 5/5 pass METHODOLOGY OK
Pass = the discipline is working. A messy report on Tier 4 data is not the failure mode — a clean report that fabricates to fill coverage gaps is.
Pass = the discipline is working. A messy report on Tier 4 data is not the failure mode — a clean report that fabricates to fill coverage gaps is.