Position Discipline Check

What this skill does

Reviews every open position with adversarial rigor. Forces you to confront the possibility that your thesis is wrong, your sizing is emotional, and your counterparty knows something you don't.

This is not a status update. This is a stress test of your decision-making.

When to use

• Weekly as a scheduled discipline review
• After any major news event that could invalidate a thesis
• When your portfolio P&L makes you feel either very good or very bad
• Before adding to any existing position
• After kill conditions trip (to evaluate whether to re-engage or close out)

Prerequisites

npm install -g @spfunctions/cli
sf login

Open positions are user-scoped.

Instructions

Step 1: Gather all data

Run these tools in parallel:

sf me portfolio --json --limit 20             # all current positions with P&L
sf portfolio history --json --ticks 10 --since 2026-04-01  # recent decisions
sf me theses --json --limit 10                 # active thesis state + confidence
sf feed --json --hours 168                     # last 7 days of signals

For each open position, also:

sf inspect <ticker> --json                     # current market state

(Batch this: parallel-fetch for all your position tickers.)

Step 2: Per-position review

For EACH open position, work through these five checks. Use the heading template for each position.

═══════════════════════════════════════════
POSITION: <ticker> | qty <n> @ avg <price> | current <price> | P&L $<n>
═══════════════════════════════════════════

#### Check 1: Entry Hypothesis

State in one sentence: why did you enter this position? What was the core belief?

If you cannot state it clearly, that itself is a finding. Note "ENTRY HYPOTHESIS UNCLEAR" and consider closing.

#### Check 2: Falsification Condition

What specific, observable event would prove your entry hypothesis wrong?

Not "things could change" — a concrete condition.

Good: "If OPEC announces 2M bpd production increase, the supply shock thesis breaks." Bad: "If the situation changes, I'll reconsider."

If no falsification condition was defined at entry, define one now and record it in the thesis (sf signal <thesis-id> "Falsification: ...").

Then check: has this condition been met or partially met? Scan the sf feed output for relevant signals. If yes, the position should probably close.

#### Check 3: From-Zero Test

Forget you own this position. If you were evaluating it fresh today:

• Would you buy it at the current price?
• At what size?
• Is it in the top-3 trades available right now? (Compare to sf discover --quality --json --limit 10)

If the answer is "no I wouldn't buy this fresh", then why are you still holding it? Sunk cost is not a reason. Either close or articulate the new reason for holding.

#### Check 4: Counterparty Read

Someone is on the other side of every trade. What do they think they know that you don't?

Examine:

• Volume profile (sf inspect <ticker> --json | jq .indicators.vr) — is there elevated volume against you?
• Cross-venue counterpart price — is the other venue priced against you?
• Recent feed signals — has new information arrived that you under-weighted?

If you can't articulate the bear case from a smart counterparty's perspective, you may be missing something.

#### Check 5: Sizing Emotion Check

Is the current size driven by:

• Your original Kelly calculation? ✓
• "I want to recover prior losses"? ✗
• "It's working, let me add"? ✗ (this is FOMO disguised as conviction)
• "It's losing, let me average down"? ✗ (this is the most expensive cognitive bias)

If size has drifted from the original Kelly fraction, document why. Usually the right move is to scale back to discipline.

Step 3: Cross-position correlation

After per-position review, check the portfolio level:

sf me portfolio --json | jq '.positions[] | {ticker, category, exposure}'

Group positions by category. Are you accidentally concentrated?

• Is the macro book all-long inflation?
• Are the geopolitical positions all-long Iran risk?

Concentrated bets have correlated downside. A single event can wipe a whole category. If any category is >40% of book, you've made a bigger bet than you intended.

Step 4: Produce the audit output

Structure:

## Discipline Audit — <date>

### Summary
- Open positions: <n>
- Total exposure: $<n>
- Daily P&L: $<n>  |  Total: $<n>

### Action items
- CLOSE: <ticker> — <reason: entry hypothesis broken | falsification hit | from-zero fail>
- REDUCE: <ticker> by <pct> — <reason: size drift | concentration concern>
- HOLD: <ticker> — <reason: hypothesis intact, sizing disciplined>
- ADD (if you have a separate <ticker>-specific case): — <reason>

### Concerns to monitor
- <position>: <concern>
- <position>: <concern>

### Category concentration
- macro: <pct>%
- geopolitics: <pct>%
- crypto: <pct>%

### Cognitive bias flags
- "I'm holding <ticker> because I don't want to admit the loss" — yes/no
- "I'm adding to <ticker> because it's been working" — yes/no
- "I'm not closing <ticker> because it might come back" — yes/no

Step 5: Execute actions

For each CLOSE action:

sf intent sell <ticker> <quantity> --note "Discipline audit close: <reason>" --json
sf intent approve <intentId>

For each REDUCE action:

sf intent sell <ticker> <reduction-quantity> --note "Discipline audit reduce: <reason>" --json
sf intent approve <intentId>

Document the action chain in the linked thesis:

sf signal <thesis-id> "Audit <date>: <action> on <ticker> — <reason>"

Bias avoidance

The hardest check in this skill is being honest about disposition effect (holding losers, selling winners). Watch for these signs:

• You spend more words rationalizing a losing position than a winning one
• You can articulate the bear case for winners but not for losers
• You say "let it bounce" or "give it time" — these are non-answers

If you catch yourself doing any of these, the position is probably the close action.

Related skills

• sf-pre-trade-check — adversarial check for NEW entries (before opening)
• sf-portfolio-review — descriptive audit (what happened, not what to change)
• sf-thesis-monitor — set up automatic kill conditions to enforce future discipline

Reference docs

• <https://docs.simplefunctions.dev/concepts/risk-gates>
• <https://docs.simplefunctions.dev/cli/portfolio>
• <https://docs.simplefunctions.dev/build/trade-intents>