coco-research/brain:wiki
Browse, search, and generate Wikipedia-style knowledge articles from the brain. Auto-generated from brain DB entities, emails, docs, and decisions.
npx skills add https://github.com/coco-research/coco --skill brain:wikiBrowse, search, and generate Wikipedia-quality articles about people, systems, teams,
and org units across all CoCo brain DB projects.
Articles are auto-generated from the brain DB by the daily knowledge engine cron
(~/.coco/knowledge/cron.py). Each article synthesizes all evidence available about
an entity — decisions, events, relationships, tasks — into a structured, versioned
knowledge artifact.
/brain-wiki [entity] — Show article (or list all)With entity name: Look up and display a knowledge article.
python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki --name "{entity}"
Display format:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
{Title} [{type}]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Infobox: Projects: {list} | Role: {role} | Team: {team}
{Summary paragraph}
## Role
{content}
## Relationships
{content}
## Timeline
{content}
## Decisions
{content}
## Open Questions
{content}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Generated: {date} · Confidence: {0-100}% · Sources: N chunks
GID: {uuid} · v{version}
If article not found → run wiki-search with the entity name as query and show top 3 candidates:
No article found for "{entity}". Did you mean:
1. {title} ({confidence}%) — /brain-wiki {gid}
2. {title} ({confidence}%)
3. {title} ({confidence}%)
Warnings to show inline:
⚠ Low confidence — this article needs more source data. Run /brain-update after adding more context.ℹ Possible duplicate: matches "{other_name}" ({similarity}%) — run /brain-wiki review-merges to resolveWithout entity name: List all available articles.
python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki-search "*" --limit 50
Display as table:
KNOWLEDGE BASE · N articles
═══════════════════════════════════════════════════════════
Name Type Confidence Updated
─────────────────────────────────────────────────────────
Alice Example person 87% 2h ago
VendorPortal system 91% yesterday
DataPipeline system 76% 2d ago
Platform Team team 82% 2d ago
...
/brain-wiki search <query> — Unified FTS5 + semantic searchpython3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki-search "{query}"
Runs both FTS5 keyword search and MemPalace semantic search, merges via RRF.
Display format:
SEARCH: "{query}" · N results · FTS5 + semantic
══════════════════════════════════════════════════════════════
# Name Type Confidence Projects Updated
─────────────────────────────────────────────────────────────────
1 {title} {type} {conf}% {projects} {time}
2 ...
Run /brain-wiki {name} to read the full article.
If 0 results:
No articles found for "{query}".
Entity may not be in any brain DB yet, or articles haven't been generated.
Run: /brain-wiki generate to generate articles for all brain entities.
/brain-wiki generate [project] — Generate / refresh articlesGenerates or refreshes knowledge articles from brain DB evidence.
Procedure:
Generate articles for project "{project}"?
This will use Claude API (estimated $0.XX for N entities).
[Y/n]
If no project specified, confirm all:
Generate articles for ALL registered projects?
Registered: {slug1}, {slug2}, ... (N projects, ~N entities)
Estimated cost: $0.XX · Estimated time: N minutes
[Y/n]
# Single project
python3 ~/.coco/knowledge/cron.py --run --project {slug} --phases 2,3,5
# All projects
python3 ~/.coco/knowledge/cron.py --run --phases 2,3,5
Phase 2: Harvesting evidence...
✓ {project}: N entities, N evidence chunks
Phase 3: Generating articles...
✓ Generated: Alice Example (confidence: 87%)
✓ Generated: VendorPortal (confidence: 91%)
~ Skipped: 3 entities (unchanged)
Phase 5: Indexing...
✓ N articles indexed (FTS5)
─────────────────────────────────────────
KNOWLEDGE ENGINE
================
Articles generated: N new, N updated
FTS5 indexed: N
Estimated cost: $0.XXX
Articles written to: ~/.coco/knowledge/articles/
Search with: /brain-wiki search "{project}"
--force flag to regenerate all articles regardless of staleness./brain-wiki people — Show cross-project people graphShow all person-type articles and cross-project relationship statistics.
Procedure:
import sys, sqlite3, json
sys.path.insert(0, str(Path("~/.coco/knowledge").expanduser()))
from schema import KNOWLEDGE_DB_PATH
conn = sqlite3.connect(KNOWLEDGE_DB_PATH)
# People articles
people = conn.execute("""
SELECT ge.canonical_name, ge.aliases_json, ge.merged_from_json,
a.confidence, a.generated_at, a.version
FROM global_entities ge
LEFT JOIN articles a ON a.gid = ge.gid
WHERE ge.type = 'person'
ORDER BY a.confidence DESC NULLS LAST
""").fetchall()
# Pending merges
merges = conn.execute("""
SELECT COUNT(*) FROM cross_project_connections
WHERE connection_type = 'proposed_merge'
""").fetchone()[0]
# Works-with edges
edges = conn.execute("""
SELECT COUNT(*) FROM cross_project_connections
WHERE connection_type = 'works_with'
""").fetchone()[0]
PEOPLE GRAPH · N people across N projects
════════════════════════════════════════════════════════
Name Projects Confidence Updated
─────────────────────────────────────────────────────────
Alice Example my-project 87% 2h ago
...
Works-with relationships: N edges inferred
Pending merge proposals: N — run /brain-wiki review-merges to resolve
No people articles found. Run /brain-wiki generate to build the knowledge base.
/brain-wiki review-merges — Review and approve entity merge proposalsInteractively review merge proposals (entities that may be duplicates).
Procedure:
python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki --list-merges
MERGE PROPOSAL (similarity: 92%)
══════════════════════════════════════════════════════════
KEEP candidate A: KEEP candidate B:
───────────────────────────── ─────────────────────────
Name: Alice Example Name: Alice Examplé
GID: {uuid-a} GID: {uuid-b}
Projects: project-a, project-b Projects: vendor-integration
Summary A: {2-3 sentences} Summary B: {2-3 sentences}
══════════════════════════════════════════════════════════
Action: [A=keep A | B=keep B | s=skip | q=quit]
python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki --approve-merge {gid_keep} {gid_remove}
This:
DELETE FROM articles_fts WHERE gid=gid_remove then re-INSERT with gid_keep
MERGE REVIEW COMPLETE
=====================
Approved: N merges
Skipped: N pairs
Remaining: N pending
/brain-wiki install-cron — Set up daily improvement cronInstall the daily knowledge engine job via macOS launchd.
Procedure:
python3 ~/.coco/knowledge/cron.py --install
KNOWLEDGE CRON INSTALLED
========================
Schedule: Daily at 02:00
Claude binary: {resolved path}
Python binary: {resolved path}
Plist: ~/Library/LaunchAgents/com.coco.knowledge-cron.plist
Log: ~/.coco/knowledge/cron.log
To uninstall: /brain-wiki uninstall-cron
To test now: python3 ~/.coco/knowledge/cron.py --run --dry-run
_find_claude_binary() with install instructions.To uninstall:
python3 ~/.coco/knowledge/cron.py --uninstall
/brain-wiki stats — Show knowledge engine statisticsDisplay current state of the knowledge engine without running any generation.
Procedure:
Query knowledge.db directly:
import sqlite3, json
from pathlib import Path
conn = sqlite3.connect(Path("~/.coco/knowledge/knowledge.db").expanduser())
stats = {
"entities": conn.execute("SELECT COUNT(*) FROM global_entities").fetchone()[0],
"articles": conn.execute("SELECT COUNT(*) FROM articles").fetchone()[0],
"fts_rows": conn.execute("SELECT COUNT(*) FROM articles_fts").fetchone()[0],
"pending_merges": conn.execute(
"SELECT COUNT(*) FROM cross_project_connections WHERE connection_type='proposed_merge'"
).fetchone()[0],
"works_with": conn.execute(
"SELECT COUNT(*) FROM cross_project_connections WHERE connection_type='works_with'"
).fetchone()[0],
"last_gen": conn.execute(
"SELECT MAX(run_at) FROM generation_log WHERE phase='3_generate' AND status='ok'"
).fetchone()[0],
"last_sync": conn.execute(
"SELECT MAX(run_at) FROM generation_log WHERE phase='6_sync' AND status='ok'"
).fetchone()[0],
"by_type": conn.execute(
"SELECT type, COUNT(*) FROM global_entities GROUP BY type"
).fetchall(),
}
Display format:
KNOWLEDGE ENGINE STATS
══════════════════════════════════════════════════
Entities: N (person: N, system: N, team: N ...)
Articles: N (FTS5 indexed: N)
Relationships: N works_with edges
Pending merges: N proposals awaiting review
Last generation: {time ago}
Last MemPalace sync: {time ago}
Articles directory: ~/.coco/knowledge/articles/ (N files)
DB size: {KB/MB}
══════════════════════════════════════════════════
Commands: /brain-wiki generate · /brain-wiki search · /brain-wiki review-merges
If knowledge.db does not exist:
Knowledge engine not initialized.
Run /brain-wiki generate to bootstrap, or /brain-wiki install-cron for daily automation.
Score normalization: score = 1.0 / (1.0 + abs(rank)) — higher score = more relevant.
(FIX M5 from review findings.)
section["content"] from each section in body_json before inserting into articles_fts.
(FIX C3 from review findings.)
DELETE WHERE gid=? then re-INSERT for any gid change (e.g., merge approvals).
(FIX M6 from review findings.)
python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py wiki-register \
--slug {slug} --db {path/to/project_brain.db}
brain-init Step 8 calls this automatically. (FIX M1 from review findings.)
~/.coco/disabled exists, all brain-wiki commands should exit silently(consistent with CoCo kill switch convention).
Take coco-research/brain:wiki from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.