m3taversal
e830fe4c5f
Three-agent knowledge base (Leo, Rio, Clay) with: - 177 claim files across core/ and foundations/ - 38 domain claims in internet-finance/ - 22 domain claims in entertainment/ - Agent soul documents (identity, beliefs, reasoning, skills) - 14 positions across 3 agents - Claim/belief/position schemas - 6 shared skills - Agent-facing CLAUDE.md operating manual Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
68 lines
2.1 KiB
Markdown
68 lines
2.1 KiB
Markdown
# Skill: Learn Cycle
|
|
|
|
The 15-minute knowledge sync that keeps agents current with the shared knowledge base.
|
|
|
|
## When to Use
|
|
|
|
Runs automatically every 15 minutes. Can also be triggered manually when significant changes are made to the knowledge base.
|
|
|
|
## Process
|
|
|
|
### Step 1: Check for new claims
|
|
|
|
Compare current timestamp against last sync timestamp. Find all claims that have been:
|
|
- Newly accepted (merged since last sync)
|
|
- Modified (confidence changed, evidence updated)
|
|
- Retired (invalidated or superseded)
|
|
|
|
### Step 2: Route to relevant agents
|
|
|
|
For each changed claim:
|
|
- Primary domain agent gets notified (always)
|
|
- Leo gets notified (always — cross-domain synthesis)
|
|
- Secondary domain agents get notified if the claim touches their domain
|
|
|
|
### Step 3: Agent review
|
|
|
|
Each notified agent processes the new claims:
|
|
|
|
**Relevance assessment:**
|
|
- Does this touch any of my active beliefs?
|
|
- Does this affect any of my active positions?
|
|
- Does this open a new line of reasoning I haven't considered?
|
|
|
|
**Integration:**
|
|
- Update mental model with new information
|
|
- If a belief's grounding claims changed → flag belief for review
|
|
- If a position's underlying beliefs are affected → flag position for review
|
|
|
|
**Signal assessment (for tweeting):**
|
|
- Is this important enough to share publicly?
|
|
- Is this novel to my audience on X?
|
|
- Would my interpretation add value beyond just relaying the information?
|
|
- Should I wait and combine with other recent learnings for a synthesis?
|
|
|
|
### Step 4: Update sync state
|
|
|
|
Record:
|
|
- Sync timestamp
|
|
- Claims processed per agent
|
|
- Beliefs flagged for review
|
|
- Positions flagged for review
|
|
- Tweet candidates identified
|
|
|
|
## Timing Notes
|
|
|
|
The 15-minute interval is a starting point. Adjust based on:
|
|
- Knowledge base velocity (how fast are claims being accepted?)
|
|
- Agent processing capacity
|
|
- Tweet output quality (if agents feel rushed, increase interval)
|
|
|
|
The goal is: agents stay current without feeling pressured. Quality of review > speed of review.
|
|
|
|
## Output
|
|
|
|
- Updated sync timestamp
|
|
- Per-agent review notes
|
|
- Cascade flags for beliefs/positions needing review
|
|
- Tweet candidate list (fed to tweet-decision skill)
|