memory-refresh

name	memory-refresh
description	Periodic memory maintenance — archive session files, convert recurring knowledge into summaries, update the memory index. Activate at end of a major phase (Discovery complete, sprint done) or when memory spans many sessions. For emergency context-window pressure mid-task, use memory-compact instead.
license	MIT
compatibility	Works with any filesystem-based AI coding agent
metadata	{"author":"gaai-framework","version":"1.0","category":"cross","track":"cross-cutting","id":"SKILL-MEMORY-REFRESH-001","updated_at":"2026-03-03T00:00:00.000Z","status":"stable"}
inputs	["contexts/memory/index.md (registry — read first to discover all active categories)","contexts/memory/** (any category registered in index.md — resolved at runtime)"]
outputs	["contexts/memory/summaries/*.summary.md","contexts/memory/archive/**","contexts/memory/index.md (updated)"]

Memory Refresh

Purpose / When to Activate

Activate:

• After major discovery or delivery phases
• When memory exceeds comfortable context size
• Periodically on active projects (weekly or per sprint)

This skill governs and optimizes existing memory only — it does not create new knowledge.

---

Process

1. Read memory index
2. Extract durable knowledge from session memory
3. Convert recurring or validated information into summary memory
4. Archive raw session files to contexts/memory/archive/
5. Compact with R7 category-aware rules:
   • Durable memory (decisions, patterns, project, ops, contacts, domains): only entries with explicit supersession markers (> SUPERSEDED by DEC-XX, > RETRACTED, > OBSOLETE — {reason}) may be archived. All other entries are ACTIVE and MUST NOT be archived. Decisions are already individual ADR files (decisions/DEC-{N}.md) — no compaction needed. For other oversized durable files → domain-split, not archive.
   • Ephemeral memory (sessions): standard compaction — summarize and archive. 5b. Summary lifecycle — mandatory when extending summary coverage: When creating or extending a decisions summary file that supersedes an existing one (e.g., decisions-90-145.summary.md → decisions-90-155.summary.md): a. Write the new summary file with updated frontmatter (id, updated_at, source range). b. Update index.md Summaries section: replace the old filename with the new one. c. Delete the old summary file from disk (it is now superseded; do not leave orphan files). d. Verify: confirm old filename is gone from disk AND index.md pointer is updated before proceeding. Also check decisions/_log.md for any ⚠️ PENDING: extend summary range notes — process them and remove the notes once handled.
6. Update memory index: update index.md frontmatter updated_at to reflect this refresh run.

---

Quality Checks

• Active memory remains minimal and high-signal
• Summaries become the primary long-term context source
• Raw exploration is archived, not deleted
• No uncontrolled memory growth
• Index always reflects current active memory
• No active durable memory entry archived — only superseded/retracted entries may be archived (R7/R7b)

---

Non-Goals

This skill must NOT:

• Create new project memory
• Record new decisions (use memory-ingest or decision-extraction)
• Generate architecture context

Governs existing memory. Keeps it clean, cheap, and precise.