ClawAudit verdict
agent-openclaw-memory
Self-evolving memory system that uses local Qdrant vector DB and Obsidian vault sync; installs from GitHub but operates fully locally with no external data egress beyond the initial package install.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. ClawAudit flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Permission integrity
package_install
Findings (10)
Pipe to python — executes piped content as Python code
SKILL.md · prose · downgraded · | Python
Writes to SKILL.md — self-modifying skill
SKILL.md · prose · downgraded · SKILL.md — they teach the Agent to write
HTTP request to bare IP address — common in malicious payloads
scripts/compress_to_rule.py · prose · downgraded · http://127.0.0.1
subprocess execution — runs system commands from Python
scripts/extract_memories.py · prose · downgraded · subprocess.run(
Downloads executables from external URLs
scripts/setup.py · prose · downgraded · Download from https://
Opens WebSocket connection
scripts/compress.py · prose · downgraded · websocket
References agent memory files
scripts/elevate_frequent.py · prose · downgraded · memory.md
Python shutil file operation — copies/moves/deletes files
scripts/elevate_frequent.py · prose · downgraded · shutil.move(
Python urllib.request — network access
scripts/embedder.py · prose · downgraded · urllib.request
Python os.environ.get — reads environment variable
scripts/health_check_v2.py · prose · downgraded · os.environ.get(
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
package_install Is this flag fair?
Thanks — recorded.