Use when the user asks to update shared memory, initialize multi-agent memory, summarize decisions so far, identify open questions, or prepare a fresh session primer.
Installs just this skill. Get the whole plugin for auto-invocation.
⚡ How it fires
How this skill gets triggered: by you, by Claude, or both.
Fires itselfClaude auto-loads it when your prompt matches the work.
You can call itInvoke it directly when you want it.
Slash command/agent-shared-memory
👁️ Context preview
The summary Claude sees to decide when to auto-load this skill.
Use when the user asks to update shared memory, initialize multi-agent memory, summarize decisions so far, identify open questions, or prepare a fresh session primer.
📊 Stats
Stars16
Forks4
LanguagePython
LicenseMIT
📦 Ships with agent-collab-workspace
</> SKILL.md
agent-shared-memory.SKILL.md
---name: agent-shared-memory
description: Use when the user asks to update shared memory, initialize multi-agent memory, summarize decisions so far, identify open questions, or prepare a fresh session primer.
---# agent-shared-memory
The persistent blackboard between Claude session A, Codex session
B, and Gemini session C — none of which see each other's
conversation history natively.
`.coord/memory.yml` is **append-only**. Past decisions don't get
edited; they get superseded by new entries that reference what they
replace. This makes it an audit trail, not a mutable scratchpad.
## When to use
Trigger phrases:
- "Update the shared memory with `<decision>`."
- "Log this open question to shared memory."
- "What have agents decided so far on this project?"
- "What are the open questions blocking us?"
- "Initialize multi-agent shared memory for this project."
- "Give me a primer for a fresh agent session."
Not for:
- Per-task scratch (use `.ai/<agent>_task_*.md` from
`agent-task-splitter`).
- Per-round artifacts (use the round-specific files written by
reconciler / acceptance-gate).
- Project-level long-term context (use `.research/project_manifest.yml`
- Filter to: (a) entries from last 14 days; (b) entries tagged
as `principle` or `decision`; (c) entries referenced by the
current round's plan.yml
- Compose memory digest (≤ memory_digest_token_budget from
context_policy, default 1200 tokens)
- Return: digest text suitable to paste into session primer +
promotion-candidate flags (entries that could move to a long-
term principles file)
Main session uses the digest instead of reading memory.yml directly.
```
Compaction-by-promotion: if subagent flags entries as long-term
principles, run a follow-up step to move them out of `memory.yml`
into a separate principles file, keeping `memory.yml` bounded.
## Commit Boundary
Every agent boundary is a commit boundary (see global rule:
~/.claude/CLAUDE.md → "Commit Discipline for Multi-Agent Work"). This
makes multi-agent work auditable (commit log = agent log) and enables
surgical rollback via `git revert <hash>` of just one agent's commit.
**Specific to this skill**: every promotion to `.coord/memory.yml` is its own commit with message `memory: <decision-summary>`. The acceptance-gate skill diffs memory commits to detect inconsistent decisions across agents.
