SimpleFunctions Portfolio Review

When to use

• User asks for portfolio status, P&L review, or autopilot audit
• Weekly review: what did the portfolio manager do this week?
• After a kill condition fires or drawdown alert: what positions led to it?
• Pre-trade context: what's already in the portfolio?

Prerequisites

npm install -g @spfunctions/cli
sf login

Portfolio data is user-scoped — requires auth.

Instructions

Step 1: Current state snapshot

sf me portfolio --json --limit 20

Returns:

• Cash balance per venue
• Total exposure
• Daily / total P&L
• Active positions (ticker, qty, avg price, current price, P&L per position)
• Portfolio status (live, halted, paused)
• Last evaluation tick timestamp

If status: halted, the autopilot is paused — check why with sf me portfolio --json | jq .haltedReason.

Step 2: Recent activity

sf portfolio history --json --ticks 20 --trades 20 --since 2026-04-01

Returns the last N ticks + trades since the date:

• Ticks: each LLM evaluation cycle — handoff note (what the LLM was told), actions proposed, risk gates applied, decisions made
• Trades: each actual order — fill price, quantity, venue, linked thesis (if any)

Pull more if needed:

sf portfolio tick <tickId> --json    # full detail of one tick
sf portfolio trade <tradeId> --json  # full detail of one trade

Step 3: Strategy + view config

sf portfolio view list --json
sf portfolio strategy list --json

Views = saved investment universes (e.g., "macro events with σ>30¢ window 30d"). Strategies = playbooks the autopilot applies to view results.

If either is empty, the autopilot is running with defaults. Document this.

Step 4: Cloud autopilot state

sf portfolio cloud status --json

Returns:

• Cloud manager status (running, idle, error)
• Last successful tick time
• Next scheduled tick
• BYOK credential connection status
• Heartbeat health

Step 5: Reconstruct the week

Produce a structured summary:

1. PnL attribution For each tick → trade → P&L: which thesis or strategy generated the entry?

2. Decisions made Per tick, what did the LLM choose to do?

• Entries: how many, average size
• Exits: how many, by reason (target hit, kill condition, time decay)
• No-actions: % of ticks where nothing was changed

3. Risk gate trips How many ticks hit risk gates (max exposure, max concentration, max daily loss)? Each gate trip is in the tick's riskGates field.

4. Anomalies

• Stale data warnings (syncPositions mismatch)
• Missing handoff notes (possible bug)
• Empty action sets (autopilot saw nothing actionable)

Step 6: Output the review

Structure:

## Portfolio summary (as of <date>)

Cash:    $<n>
Exposure: $<n>
P&L:     Daily $<n>  |  Total $<n>
Status:  <live|halted|paused>

## Recent activity (last <N> days)

- <N> ticks, <N> trades, <N> kill condition fires
- Top P&L contributors: <position> +$<n>, <position> -$<n>
- Strategies: <list>
- Risk gates tripped <N> times (top reason: <reason>)

## Anomalies / things to address

- <issue 1>
- <issue 2>

## Next actions

- <action 1>
- <action 2>

Common patterns

"Why did the autopilot do X?"

# Find the tick that contains the decision
sf portfolio history --json --ticks 50 --since <date> | jq '.ticks[] | select(.actions[]?.ticker == "<ticker>")'
# Then inspect the tick
sf portfolio tick <tickId> --json

The tick has handoffNote (what the LLM was given) and decisions (what it chose), which together explain the reasoning.

"Are positions in sync between local and server?"

sf setup --sync-positions
sf status --json | jq .positionSyncState

If positionSyncState.drift > 0, run sf setup --sync-positions to push local exchange state to server context. The autopilot reads server-side, so drift causes wrong decisions.

Errors and recovery

• `status: halted`: portfolio auto-halted (drawdown, error, manual). Check haltedReason. Resume with sf portfolio resume.
• Empty `ticks` array: autopilot may not be enabled. Use sf-autopilot-setup to enable.
• Tick has `evalError`: LLM eval failed for that tick. Check evalError.message — often budget exhaustion or model rate limit.

Related skills

• sf-autopilot-setup — initial autopilot configuration
• sf-discipline-audit — adversarial review of open positions
• sf-thesis-monitor — for thesis-linked monitoring config

Reference docs

• <https://docs.simplefunctions.dev/cli/portfolio>
• <https://docs.simplefunctions.dev/api-reference/portfolio>
• <https://docs.simplefunctions.dev/guides/portfolio-autopilot>