feat(skills): add /add-mnemon skill — persistent semantic memory

[ira-at-work]

Summary

• New /add-mnemon skill that installs the mnemon persistent memory CLI into agent containers
• Agents get a queryable knowledge graph that survives container restarts and context compaction
• Each fact (insight) has a category, importance score, entity tags, and typed edges to related facts (causal, semantic, temporal, entity)
• Agents recall relevant past knowledge before responding and store new insights after each turn
• Per-agent-group data directory — no cross-group memory leakage

Test plan

• Run skill — verify mnemon binary installed in container and data directory created
• Send a message, then ask the agent to remember a fact — verify it appears in mnemon recall
• Restart the container — verify the fact persists on next session
• Verify recall is invoked before agent responds (check container logs for mnemon recall)
• Verify separate agent groups have separate mnemon stores

🤖 Generated with Claude Code

[ira-at-work]

Closing — incorrect base, will re-open with clean branch.

[gavrielc]

Thanks @ira-at-work — merging this in! Great addition to the skill library.