remember

name	remember
description	Persist important information from this conversation to long-term memory. Use proactively whenever the user mentions a person, project, company, decision, meeting, deadline, preference, or any fact that future sessions should know about. Also use when the user says remember this, store this, note this, keep track of, or don't forget.
allowed-tools	Bash(kyberbot remember *)

Remember

Stores information from terminal sessions into the brain's full memory pipeline — the same one used by Telegram, WhatsApp, and heartbeat. Without this skill, terminal conversations vanish when the session ends.

The kyberbot remember command feeds the same storeConversation() pipeline as messaging channels:

• Timeline — temporal event index
• Entity Graph — people, companies, projects, and their relationships
• Embeddings — semantic search (when ChromaDB is available)

When to Fire

Fire this skill proactively — don't wait for the user to say "remember this." If they mention a person, a decision, a project update, a preference, or any fact that future sessions would benefit from knowing — store it.

Always store when:

• The user mentions a new person and their role/relationship
• A decision is made about a project, tool, or approach
• Meeting notes or conversation summaries come up
• The user shares facts about themselves, their work, or their goals
• Deadlines, milestones, or schedule changes are discussed
• New projects or initiatives are mentioned

Don't store:

• Trivial back-and-forth ("thanks", "ok", "got it")
• Purely mechanical requests ("format this code", "fix the typo")
• Information already stored in this session

How to Store

Step 1: Compose the Memory

Summarize the key information in a clear, factual sentence. Include names, dates, and context. The text should be understandable out of context — a future session reading this should immediately grasp what happened.

Good: "Met with Sarah Chen from Notion on Feb 23 to discuss API integration for the dashboard project" Bad: "Had a meeting about stuff"

Step 2: Run the Command

kyberbot remember "<text>"

If there's a natural response or additional context to pair with it:

kyberbot remember "<text>" --response "<context>"

Step 2b: Tag the Memory (when context is clear)

When the user's message tells you the memory belongs to a specific project, has obvious sensitivity, or is cross-cutting, tag it. ARP scope policies use these tags as the source of truth for what gets shared with paired peer agents — an untagged memory is invisible to a project-scoped peer query, a tagged one is matched.

Flag	When to set	Example
--project <slug>	The memory is specifically about a named project, product, or initiative	--project alpha, --project kyberco-launch
--tag <name> (repeatable)	Cross-cutting themes the user has used; client/team names that aren't the primary project	--tag launch --tag draft
--classification <tier>	Content is sensitive	--classification pii (SSNs, addresses, health), --classification confidential (internal-only), --classification internal (default for company info), --classification public (already-public info)

Pick slugs for --project (lowercase, dashes/underscores) — they need to match what's typed in the ARP scope picker on the cloud side. "Project Alpha" → alpha. "Q2 Launch" → q2-launch.

Don't make up tags. Only set --project / --tag when the user has clearly named a project or theme — guessing from context risks mis-scoping and a peer agent seeing or missing the wrong memories.

Step 3: Confirm

Briefly acknowledge to the user that the information has been stored. A simple "Noted." or "Stored." suffices unless the user explicitly asked you to remember something, in which case confirm what you stored.

Examples

Person mentioned (no project context): User says "I talked to Jake from the infra team about migrating to Kubernetes"

kyberbot remember "Talked to Jake from the infra team about migrating to Kubernetes"

Decision in a named project: User says "For project alpha, let's go with Next.js for the frontend"

kyberbot remember "Decision: using Next.js for the frontend" \
  --response "Chosen over Remix and SvelteKit" \
  --project alpha

Meeting notes scoped to a project: User shares detailed Q2 launch meeting notes

kyberbot remember "Weekly sync with product team — discussed Q2 roadmap, prioritized auth overhaul and dashboard redesign" \
  --response "Auth overhaul starts March 1, dashboard redesign in April. Sarah leading auth, Mike on dashboard." \
  --project q2-launch \
  --tag roadmap

Sensitive content: User pastes a client SSN or contract terms

kyberbot remember "Acme Corp contract: $250K/year, auto-renews 2027-01-01" \
  --project acme-deal \
  --classification confidential

PII (highest sensitivity tier): User mentions a person's home address or health status

kyberbot remember "Sarah's home address is 123 Maple St" \
  --classification pii

Correction Detection

When the user says things like:

• "That's wrong about [entity]"
• "Actually, [entity] works at [X], not [Y]"
• "No, [correct fact]"
• "Forget that about [entity]"
• "[Entity] doesn't work at [X] anymore"

Treat this as a correction:

1. Run kyberbot recall "<entity>" to see what you currently know
2. Store the correct fact with kyberbot remember — the contradiction detection system will automatically supersede the old, lower-confidence fact
3. Confirm briefly: "Corrected."

The memory system uses source confidence weighting — facts stored via terminal remember get 0.95 confidence (user-direct), which is higher than chat messages (0.85) or AI-extracted facts (0.60). So a correction stored here will naturally take precedence over earlier, less-reliable information.

If the user says something is wrong but doesn't provide a replacement (e.g., "That's not true about John"), acknowledge the issue and ask what the correct information is.

Notes

• This skill complements (not replaces) updating USER.md, SOUL.md, and brain/ files. Use those for structured, long-lived information. Use remember for capturing the stream of events and facts.
• Memories are searchable via kyberbot recall, kyberbot timeline, and kyberbot search.