/trace - Unified Discovery System

Find + Log + Dig + Distill

Usage

code

/trace [query]              # --smart (default): Oracle first, auto-escalate
/trace [query] --oracle     # Oracle only (fastest)
/trace [query] --deep       # Full 5 parallel subagents
/trace [url]                # Auto-clone GitHub URL, then trace locally

Step 0: Timestamp

bash

date "+🕐 %H:%M %Z (%A %d %B %Y)"

URL Detection & Auto-Clone

If input is a GitHub URL, clone AND symlink first:

bash

# Replace [URL] with actual URL
ghq get -u [URL] && \
  GHQ_ROOT=$(ghq root) && \
  OWNER=$(echo "[URL]" | sed -E 's|.*github.com/([^/]+)/.*|\1|') && \
  REPO=$(echo "[URL]" | sed -E 's|.*/([^/]+)(\.git)?$|\1|') && \
  mkdir -p "ψ/learn/$OWNER" && \
  ln -sf "$GHQ_ROOT/github.com/$OWNER/$REPO" "ψ/learn/$OWNER/$REPO" && \
  echo "✓ Symlinked: ψ/learn/$OWNER/$REPO"

Verify: ls -la ψ/learn/

Then trace using ψ/learn/[owner]/[repo] path.

Note: Grep tool doesn't follow symlinks. Use Bash: rg -L "pattern" ψ/learn/

Mode 1: --oracle (Oracle Only)

Fastest. Just Oracle MCP, no extension.

code

oracle_search("[query]", limit=15)

Display results and done. Even if empty.

Mode 2: --smart (Default)

Oracle first → auto-escalate if results < 3

Step 1: Query Oracle first

code

oracle_search("[query]", limit=10)

Step 2: Check result count

•If Oracle results >= 3 → Display and done
•If Oracle results < 3 → Auto-escalate to --deep mode

Mode 3: --deep (Explore Subagents)

Launch 5 parallel Explore agents (Haiku) for thorough search.

Agent	Searches
1	Current repo files
2	Git history (commits, creates, deletes)
3	GitHub issues
4	Other repos (ghq, ~/Code)
5	Retrospectives & learnings (ψ/memory/)

Use Task tool with subagent_type="Explore" for each agent

After search, auto-log to Oracle:

code

oracle_trace({
  query: "[query]",
  foundFiles: [...],
  foundCommits: [...],
  foundIssues: [...]
})

Trace Logging

Always log traces to: ψ/memory/traces/YYYY-MM-DD-HHMM-query-slug.md

bash

# Create traces directory
mkdir -p "$ROOT/ψ/memory/traces"

# Generate filename
# Example: 2026-01-23-1430-claude-code.md

Write trace file with this format:

markdown

---
query: "[query]"
mode: [oracle|smart|deep]
timestamp: YYYY-MM-DD HH:MM
oracle_results: [count]
escalated: [true|false]
---

# Trace: [query]

**Mode**: [mode]
**Time**: [timestamp]

## Oracle Results
[list results or "None"]

## Local Files
[list files found or "None"]

## Git Commits
[list commits or "None"]

Philosophy

Trace → Dig → Trace Deeper → Distill → Awakening

The Seeking Signal

User Action	Meaning	AI Response
`/trace X`	First search	--smart (Oracle first)
`/trace X` again	Still seeking	Oracle knows
`/trace X --deep`	Really need it	Go deep with subagents
Found!	RESONANCE	Log to Oracle

Auto-Escalation Flow

code

/trace [query]     → Oracle search (what we know)
      ↓
  < 3 results?     → Auto-escalate to --deep
      ↓
/trace --deep      → 5 subagents explore everywhere
      ↓
  FOUND!           → 🔮 RESONANCE! Log to Oracle
      ↓
  Next session     → Easier to find (knowledge extended)

Summary

Mode	Speed	Scope	Auto-Escalate
`--oracle`	Fast	Oracle only	No
`--smart`	Medium	Oracle → maybe deep	Yes (< 3 results)
`--deep`	Thorough	5 parallel agents	N/A

ARGUMENTS: $ARGUMENTS