Agent Skills
› winstonkoh87/Athena-Public
› decision-journal
decision-journal
GitHub统一决策生命周期管理技能，涵盖事前记录、事后复盘、失败分类及校准追踪。支持吸收旧版后记引擎，通过标准化模板辅助用户客观评估假设、区分运气与执行失误，并长期跟踪置信度校准情况。
Trigger Scenarios
做出决定前需记录
回顾过去决策的效果
分析失败原因或进行事后复盘
检查个人判断的校准程度
Install
npx skills add winstonkoh87/Athena-Public --skill decision-journal -g -y
SKILL.md
Frontmatter
{
"name": "decision-journal",
"model": "default",
"auto-invoke": false,
"description": "Unified decision lifecycle: Pre-decision logging, post-decision review, failure classification, and calibration tracking. Absorbs: post-mortem-engine.",
"allowed-tools": [
"Write",
"Read"
],
"argument-hint": "log decision | review decisions | calibration check | post mortem | what went wrong | failure analysis",
"context_trigger": "decision, should I, trade-off, regret, post-mortem, what went wrong, calibration, prediction, hindsight, journal"
}
Decision Engine (Journal + Post-Mortem)
Absorbs:
post-mortem-engine
Complete decision lifecycle in one skill: record decisions BEFORE outcomes are known, review them AFTER, classify failures objectively, and track calibration over time.
Triggers
"I've decided to", "logging a decision", "was that a good decision", "calibration", "what went wrong", "post mortem", "failure analysis", "AAR", "I screwed up"
Part 1: Pre-Decision Entry (BEFORE outcome)
## Decision Entry: [YYYY-MM-DD HH:MM]
### The Decision
[What am I deciding to do?]
### The Alternatives
1. [Alternative A and why I rejected it]
2. [Alternative B and why I rejected it]
### My Confidence
[X]% confident this is the right call.
### Key Assumptions (numbered)
1. [Assumption 1]
2. [Assumption 2]
### What Would Change My Mind
[Specific observable evidence that would make me reverse]
### Expected Outcome
- Best case: [description] (probability: X%)
- Most likely: [description] (probability: X%)
- Worst case: [description] (probability: X%)
### Decision Class
- [ ] Reversible (Type 2 — decide fast, adjust later)
- [ ] Irreversible (Type 1 — decide carefully, no undo)
Part 2: Post-Decision Review (30-90 days later)
## Review: [Original Decision Date]
### Actual Outcome
[What actually happened?]
### Assumptions Audit
1. [Assumption 1]: [Correct / Wrong / Partially correct]
2. [Assumption 2]: [Correct / Wrong / Partially correct]
### Calibration
- Stated confidence: X%
- Would I make the same decision with same info? [Yes / No]
- Outcome due to: [good decision / luck / bad decision / bad luck]
Part 3: Post-Mortem (When Things Go Wrong)
Phase 1: Just the Facts (No Interpretation)
Timeline of observable events only. No opinions, no "I should have."
Phase 2: Root Cause (The 5 Whys)
1. Why did [outcome] happen? → Because [cause 1]
2. Why? → Because [cause 2]
3. Why? → Because [cause 3]
4. Why? → Because [cause 4]
5. Why? → Because [ROOT CAUSE]
Phase 3: Classification
| Category | Question |
|---|---|
| Process Failure | Followed system, it failed → Update system |
| Execution Failure | Deviated from system → Discipline issue |
| Information Failure | Critical info unavailable → Update model, not system |
| Luck Failure | Within expected failure rate → Change NOTHING |
Critical Rule: Luck failures do NOT get process changes. At 60% WR, 40% of trades WILL fail. Changing your process after a luck failure is the #1 way to destroy a working edge.
Output
Post-Mortem Report: [Event]
─────────────────────────────
Root Cause: [One sentence]
Classification: [PROCESS / EXECUTION / INFORMATION / LUCK]
Required Changes: [Specific actions or "NONE — within expected parameters"]
Calibration Tracking
Over 20+ reviewed decisions:
| Stated Confidence | Actual Correct % | Calibration |
|---|---|---|
| 90% | XX% | [Over / Under / Well-calibrated] |
| 70% | XX% | [Over / Under / Well-calibrated] |
| 50% | XX% | [Over / Under / Well-calibrated] |
Storage: .context/memories/decision_journal/
Integration
- Triggers
trading-risk-gateon Type 1 (irreversible) decisions - Feeds into
trade-journal-analyzerfor trading decisions - Triggers
circuit-breakerif process failures are recurring
Version History
-
e624e2d
Current 2026-07-19 08:47
整合了 post-mortem-engine 功能，提供完整的决策生命周期支持。
