- SOUL.md: full agent identity, operating principles, voice - IDENTITY.md: runtime identity, hosts, boundaries - USER.md: operator context imported from hermes-soul - AGENTS.md: actual operating rules, infrastructure, quick reference - memories/curated/: 5 topics (tailscale, forgejo, agents, projects, vaultwarden) - skills/: 9 cross-harness skills imported from hermes-soul after review - docs/PLAN-CONFIGURE-PRIVATE-REPO.md: configuration plan - Validate: passes clean
1.3 KiB
1.3 KiB
Curated Memories
Cross-harness memory summaries reviewed and imported from private runtime backups. Each file is a sanitized summary of durable knowledge that matters across harnesses.
Current memories
| File | Source | Content |
|---|---|---|
tailscale-network.md |
hermes-soul memory | Host IPs, roles, connectivity rules |
forgejo-infrastructure.md |
hermes-soul memory | Repo layout, conventions, API patterns |
agent-roster.md |
hermes-soul memory | Who runs where, multi-agent setup |
project-structure.md |
hermes-soul memory | Clawdie repos, Colibri crates, dev rules |
vaultwarden-secrets.md |
hermes-soul memory | Secrets infrastructure, bw CLI setup |
Adding memories
Run the private-source planner to discover candidates:
python3 scripts/layered_soul.py plan-private-source \
examples/private-sources/hermes-soul.example.json \
--source-root ~/hermes-soul
Review the reported candidate_paths.general_memory and candidate_paths.user_memory files. Manually summarize or redact private details, then save sanitized versions here.
Rules:
- No raw session content
- No API keys, tokens, or passwords
- No platform-specific config (Tailscale IPs are acceptable;
.envcontents are not) - Keep files focused — one topic per file, 300-1500 words