session-analyzer

name	session-analyzer
description	Analyze and dump OpenCode sessions from any project. Use when the user asks to read a session, find a session by title, list sessions for a project, dump session content, or show usage stats (skills, tools, tokens, cost) for any project directory.
metadata	{"author":"emergent","version":"1.0"}

Skill: session-analyzer

Read, search, and dump OpenCode sessions from any project on this machine.

---

Script location

/root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py

---

Commands

List sessions for a project

python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --list [N]

• Omit N to list all sessions (most recent first)
• Add --since YYYY-MM-DD to filter by date
• Add --search "query" to filter by title substring

Find a session by title

python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --search "visibility of memory"

Find a session by content keyword (when title is unknown)

When the user gives you a keyword from the conversation (not the title), use --content-search. This searches the actual message text across all sessions for the project:

python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --content-search "APIEndpoint"

Returns matching sessions with the surrounding context snippet. Use this when --search (title-only) returns nothing.

Dump a session (to stdout or file)

# By session ID
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --session ses_2f374b1c2ffeS3zARC4vnePW3Q

# By title substring (auto-resolves if unique)
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --session "visibility of memory"

# Save to file
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --session "visibility of memory" \
  --output /tmp/session-dump.md

Usage stats for a project

# All-time stats
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --stats

# Last 30 sessions
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --stats --sessions 30

# Since a date
python3 /root/emergent.memory/.opencode/skills/session-analyzer/session_dump.py \
  --project /path/to/project \
  --stats --since 2026-03-01

---

Workflow

1. Identify the project path from the user's request (e.g. /root/legalplant-api, /root/emergent.memory)
2. Run the appropriate command above
3. Present results — for dumps, summarize the key conversation beats; for stats, call out top skills/tools/cost

Known project paths

Project	Path
emergent.memory	/root/emergent.memory
legalplant-api	/root/legalplant-api
twentyfirst	/root/twentyfirst
lawmatics	/root/lawmatics
specmcp	/root/specmcp
emergent (UI)	/root/emergent
doc-processing-suite	/root/doc-processing-suite
diane	/root/diane

---

Notes

• DB is at ~/.local/share/opencode/opencode.db (read-only queries)
• Project is resolved by matching the worktree column in the project table
• Session content lives in the part table (not message.data) — the script handles this correctly
• Tool calls are shown inline with [TOOL: name] / [RESULT: name] markers
• Long tool inputs/outputs are truncated at 800/1000 chars in dumps