# Outcome Memory

> Close the loop between recommendations and real outcomes so future Codex judgment improves over time.

- Skill: `mphaxise/outcome-memory` (Agent Skill)
- Install (CLI): `npx skillmds@latest add mphaxise/outcome-memory`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mphaxise/outcome-memory/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mphaxise (https://skillmd.com/u/mphaxise)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mphaxise/outcome-memory

---


# Outcome Memory

Use this skill after a plan, review, ship, design decision, startup memo, or ethics review has had enough time to produce a result.

This is a Praneet-specific hand-port enhancement. It makes judgment compound by recording whether prior advice worked, failed, or needs revision.

## Workflow

1. Identify the prior decision or recommendation.
2. Gather outcome evidence:
   - what happened
   - what shipped
   - user or stakeholder response
   - metrics or qualitative signal
   - surprises
   - harms or exclusions discovered later
3. Compare expectation to reality:
   - right for the right reason
   - right for the wrong reason
   - wrong but useful
   - wrong and misleading
   - still unknown
4. Extract learning:
   - update future scorecards
   - revise assumptions
   - add follow-up tasks
   - capture ethics or accessibility lessons
5. Persist:
   - write a report when the outcome should be reviewable
   - use `capture` or `brain-ops` for durable memory
   - route to `retro` when this is team/process learning

## Guardrails

- Do not force a lesson when evidence is still ambiguous.
- Do not hide negative outcomes.
- Separate personal preference drift from actual user or ethical evidence.