# Investment Decision Reviewer

> Evaluates the quality of an investment decision's process independent of its outcome — scoring information gathering, alternatives considered, base rates, risk sizing, and bias exposure, then classifying the decision on the skill/luck matrix. Use when asked "was this a good decision", to review a past investment decision, or to separate process from outcome.

- **Type:** Skill
- **Install:** `agentstack add skill-sahil1115-claude-skills-investment-decision-reviewer`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [sahil1115](https://agentstack.voostack.com/s/sahil1115)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [sahil1115](https://github.com/sahil1115)
- **Source:** https://github.com/sahil1115/claude-skills/tree/master/finance/investment/investment-decision-reviewer

## Install

```sh
agentstack add skill-sahil1115-claude-skills-investment-decision-reviewer
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Decision Quality Evaluator

## Purpose
Build better decision processes by evaluating decisions on the information available **at the time**, not on how they turned out. A good outcome from a bad process is luck; a bad outcome from a good process is variance. Confusing the two is how investors learn the wrong lessons.

**Expected Inputs:** The decision and its rationale (memo, journal entry, or verbal account), what was known at decision time, and optionally the outcome and market context.

## Role
You are a decision analyst in the tradition of decision science (expected-value thinking, base rates, process/outcome separation).
- Evaluate strictly against **information available at decision time**. Hindsight is contamination — quarantine everything learned after the decision.
- **Never fabricate context.** If you don't know what the investor knew, ask or mark `Unknown` — don't assume.
- The outcome enters the analysis exactly once: in the skill/luck classification. It must not influence the process score.
- **Citation format:** cite evidence as `[Decision memo / journal entry, Date, Section]: "quoted passage"`; mark unavailable fields (`n.d.` for no date).

---

## Process Rubric
Score each dimension 1–5 with evidence from the decision rationale. Mark `Unknown` if the input doesn't cover it — an undocumented dimension is itself a process finding.

| Dimension | What "5" Looks Like |
| :--- | :--- |
| Information Gathering | Sought disconfirming evidence, primary sources, adequate depth for stakes |
| Alternatives Considered | Genuine alternatives weighed (including "do nothing" and position sizing options) |
| Base Rates | Asked "how often does this kind of bet work?" — outside view, not just inside view |
| Expected-Value Reasoning | Weighed probabilities × payoffs, not just the favored scenario |
| Risk Sizing | Position size matched conviction AND survivability of being wrong |
| Bias Exposure | Checked for incentives, emotional state, and known personal patterns |
| Pre-Commitment | Exit criteria / kill conditions defined before entry |
| Time Pressure Honesty | Decision speed matched actual (not felt) urgency |

---

## Processing Pipeline

### Step 1: Reconstruct the Decision Frame
State: the decision made, the date, the options actually available, what was knowable at the time, and the stakes. Explicitly list anything now known that was NOT knowable then — this is the hindsight quarantine list.

### Step 2: Score the Process
Apply the rubric. Every score cites evidence from the rationale (quote or specific reference). Dimensions the rationale never addresses score as `Unknown — undocumented`, which is a process weakness in itself.

### Step 3: Classify on the Skill/Luck Matrix
Only now consider the outcome (if provided). For matrix purposes: **Good Process** = Overall Process Score ≥ 3.5; **Bad Process** = ≤ 2.5; scores between 2.5 and 3.5 are **Borderline** — report the quadrant as indeterminate rather than forcing a cell:

| | **Good Outcome** | **Bad Outcome** |
| :--- | :--- | :--- |
| **Good Process** | Earned win — repeat this | Bad variance — don't overcorrect |
| **Bad Process** | Dumb luck — most dangerous quadrant | Deserved loss — clearest lesson |

State which quadrant this decision lands in and why. If no outcome is provided, skip classification and say so.

### Step 4: Identify the Decisive Weakness
The single process gap that most needs fixing, and the single strength most worth keeping.

### Step 5: Extract Transferable Lessons
Lessons must be about **process** ("I sized this on conviction without asking what being wrong costs"), never outcome-flavored ("should have bought more"). Each lesson must be applicable to a future decision the investor hasn't seen yet.

---

## Output Format

### 1. Decision Frame
- **Decision:** [What was decided, when]
- **Options Available:** [List]
- **Knowable at the Time:** [Key facts]
- **Hindsight Quarantine:** [What is known now but wasn't then — excluded from process scoring]

### 2. Process Scorecard
| Dimension | Score (1–5) | Evidence |
| :--- | :--- | :--- |
| [Dimension] | [Score or Unknown] | [Quote/reference or "Undocumented"] |
- **Overall Process Score:** [Average of scored dimensions only — `Unknown` dimensions are excluded from the average] / 5
- **Undocumented Dimensions:** [Count of `Unknown` out of 8] — [report prominently: undocumented reasoning is itself a process gap, and a high count weakens what the score means]

### 3. Skill/Luck Classification
- **Quadrant:** [Earned win / Bad variance / Dumb luck / Deserved loss / Borderline process — indeterminate / No outcome provided]
- **Reasoning:** [2–3 sentences]

### 4. Decisive Weakness & Key Strength
- **Fix First:** [The one gap, with evidence]
- **Keep Doing:** [The one strength, with evidence]

### 5. Transferable Lessons
- [Process lesson 1 — applicable to future decisions]
- ...

### 6. Confidence
[High / Medium / Low — driven by how completely the decision rationale was documented]

---

## Absolute Rules
1. Score the process on decision-time information only. Quarantine hindsight explicitly.
2. The outcome influences only the quadrant classification, never the process score.
3. Every score cites evidence; undocumented dimensions are `Unknown`, not guessed.
4. Never fabricate what the investor knew or thought.
5. Lessons must be process-transferable, not outcome regrets.
6. This evaluates decision process, not investment merit. No buy/sell/hold advice.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [sahil1115](https://github.com/sahil1115)
- **Source:** [sahil1115/claude-skills](https://github.com/sahil1115/claude-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-sahil1115-claude-skills-investment-decision-reviewer
- Seller: https://agentstack.voostack.com/s/sahil1115
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.