kk-data

name: kk-data description: Access local data stores including chat logs, transcripts, agent memory, and workspace files. metadata: openclaw: requires: bins: ["python3"] env: ["KK_AGENT_NAME"]

Local data access for KarmaCadabra agents. Provides access to chat logs, transcripts, agent memory files, and workspace data stored on the agent's local filesystem.

Data Locations

Data Type	Path	Format
Chat logs	`logs/YYYYMMDD/full.txt`	Plain text, one message per line
Transcripts	`data/transcripts/YYYYMMDD/{id}/transcripcion.json`	JSON with timestamps and text
Agent memory	`data/workspaces/{agent}/memory/MEMORY.md`	Markdown
Workspace	`data/workspaces/{agent}/`	Mixed files
Reputation snapshots	`data/reputation/`	JSON snapshots

Script

read_agent_memory.py

Located at scripts/kk/read_agent_memory.py. Read another agent's MEMORY.md from their workspace directory. Useful for understanding what other agents know and have learned.

python3 scripts/kk/read_agent_memory.py --agent kk-skill-extractor
python3 scripts/kk/read_agent_memory.py --agent kk-coordinator --section "Trusted Agents"

Arguments:

--agent (required): Agent name to read memory from
--section (optional): Specific ## Section Name to extract from the MEMORY.md

Output (full memory):

{
  "agent": "kk-skill-extractor",
  "content": "# Agent Memory\n\n## Skills Database\n..."
}

Output (specific section):

{
  "agent": "kk-coordinator",
  "section": "Trusted Agents",
  "content": "## Trusted Agents\n\n- kk-karma-hello: reliable, fast\n..."
}

Output (not found):

{
  "agent": "kk-new-agent",
  "content": "",
  "note": "No MEMORY.md found at data/workspaces/kk-new-agent/memory/MEMORY.md"
}

The script tries both kk-{name} and {name} prefixes when searching.

Accessing Logs Directly

Chat logs are plain text files organized by date. Each agent can read its own logs directory:

# List available log dates
ls logs/

# Read today's logs
cat logs/20260225/full.txt

# Count messages
wc -l logs/20260225/full.txt

Accessing Transcripts Directly

Transcripts are JSON files organized by date and session ID:

# List available transcript dates
ls data/transcripts/

# List sessions for a date
ls data/transcripts/20260225/

# Read a transcript
cat data/transcripts/20260225/{session-id}/transcripcion.json

Dependencies

lib.memory (read_memory_md)

Error Handling

Exit code 1 on failure with JSON error to stderr:

{"error": "description of what went wrong", "agent": "kk-karma-hello"}