decision-loglisted
Install: claude install-skill NITISH-R-G/agent-memory-protocol
# Decision Log
**Relationship to `architecture-decision` elsewhere in this skill ecosystem**: that skill covers *authoring* a full Architecture Decision Record as a standalone practice. This skill is the same underlying discipline, specifically wired into the `.memory/` system's file layout and lifecycle — if a project already uses `architecture-decision`, point its output at `.memory/decisions/` rather than maintaining two separate decision-tracking locations.
## Why the rejected alternatives matter more than the choice
The single most valuable thing a decision record can prevent is a future session — possibly in a different tool, possibly months later — re-proposing an approach that was already tried and specifically rejected, for reasons that are no longer visible anywhere except someone's memory of a conversation that's now gone. A decision record that states only the final choice, with no record of what else was considered, doesn't actually prevent this.
## Format
```markdown
# 0012: Extract TokenService from auth middleware
## Status
Accepted
## Context
Auth middleware was handling three unrelated concerns: request validation, token refresh,
and session logging. Token refresh logic specifically needed dedicated test coverage that
was hard to get at while it was inline.
## Decision
Extract token refresh into a dedicated TokenService class, injected into the middleware
rather than implemented inline.
## Alternatives considered and rejected
- **Keep inline, extrac