Case 01479
A fast, offline static risk analysis CLI for AI agent skill files. Detects malicious instructions, steganographic payloads, and dangerous capability chains.
npx -y skills add knownasnaffy/prompthound --skill case_01479Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Migrate EasyClaw workspace-level behavior into a new OpenClaw workspace by locating and comparing old EasyClaw brain files such as AGENTS.md, SOUL.md, MEMORY.md, HEARTBEAT.md, memory/, launchd/context-management docs, then staging or importing them with backups. Use when a user wants to copy EasyClaw principles, memories, heartbeat logic, context-management setup, or other non-JSON core system behavior into the current OpenClaw workspace.
SKILL.md
3.3 KB, 681 tokens by cl100k_base, as published. Nobody here has run it
EasyClaw Brain Migration
Use this skill for the files that define the assistant's operating style and continuity, not just runtime config.
Source locations to check
Primary legacy workspace:
~/.easyclaw/workspace/
Important source files:
AGENTS.mdSOUL.mdCORE-PRINCIPLE.mdMEMORY.mdUSER.mdHEARTBEAT.mdmemory/docs/context-management.mdscripts/load_context.sh- launchd files under
~/Library/LaunchAgents/com.easyclaw.*.plist
Target workspace:
- current OpenClaw workspace root
Migration policy
Copy directly
These usually copy well when missing in the target workspace:
MEMORY.mdmemory/contents- context snapshots / historical notes that are purely informational
Merge carefully
These should be reviewed or staged before replacing active files:
AGENTS.mdSOUL.mdHEARTBEAT.mdUSER.md
Why: the new workspace may already contain OpenClaw-native conventions that should not be blown away.
Rebuild, don't blindly copy
These are automation mechanisms, not just text:
- launchd plists
- helper scripts invoked by launchd
- old context-management wrappers
Translate them into current OpenClaw mechanisms such as:
HEARTBEAT.md- cron jobs
- workspace scripts
- updated notes in
AGENTS.md/TOOLS.md
Workflow
1. Generate a migration report
Run:
python3 scripts/report_easyclaw_brain.py
This reports:
- what source files exist
- what target files exist
- which files are safe to import directly
- which should be staged for review
- legacy automations that need rebuilding
2. Stage legacy brain files safely
Run:
python3 scripts/stage_easyclaw_brain.py
This copies legacy brain files into imports/easyclaw/ inside the current workspace without overwriting active files.
Use staging first when you want a safe reviewable landing zone.
3. Import memory into active workspace
Run:
python3 scripts/stage_easyclaw_brain.py --import-memory
This:
- backs up current target files if needed
- copies
MEMORY.mdif missing - copies legacy
memory/files that do not already exist - leaves active
AGENTS.md,SOUL.md, andHEARTBEAT.mduntouched
4. Merge principles manually or with explicit review
After staging, read the staged files and selectively merge:
- stable operating rules from
AGENTS.md - persona/principles from
SOUL.mdandCORE-PRINCIPLE.md - heartbeat tasks that still make sense
- context-management notes from
docs/context-management.md
Do not auto-enable old schedules unless the user wants them restored.
Notes
- Treat old memory and prompts as sensitive workspace data.
- Prefer additive migration over destructive replacement.
- For heartbeat and scheduling logic, surface what existed and ask whether to rebuild it as cron/heartbeat behavior.