Security audit
Security checks for vulnerabilities and agentic risk
The skill’s local preference recording and temporary quiz match its disclosed purpose and include clear user controls.
Expect automatic storage of short preference excerpts in local plaintext files. Use pause to stop recording. Forget excludes an entry from future preferences but retains its original log line; delete the configured storage directory to erase saved data. Use install --apply --no-onboard to disable automatic quiz prompts. The inspected artifacts do not support the pre-scan malicious signal.
Without declared permissions the skill's intent is opaque and cannot be validated.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
---
name: xtaste
description: "Remember your taste once — every AI after that knows it. Learns the user's personal taste (visual design, writing, naming, code style, way of working, product trade-offs) from a 5-minute local quiz plus their later reactions, and applies it. Use before designing UI, writing copy, naming things or making aesthetic/product choices for the user, and whenever the user reacts to an output with likes, dislikes or corrections. Once installed it acts automatically: on first use it opens a local quiz page in the browser (once; not again for 7 days if ignored), and it records the user's taste statements to ~/.xtaste without asking each time (`xtaste pause` stops recording). Optional and only on the user's request: importing low-weight hints from a local Dongwo (懂我) prompt inbox."
metadata:
openclaw:
requires:
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
return "zh" if value.lower().startswith("zh") else "en"
if sys.platform == "darwin":
try:
out = subprocess.run(["defaults", "read", "-g", "AppleLanguages"],
capture_output=True, text=True, timeout=2).stdout
first = re.search(r'"?([A-Za-z]{2})', out)
if first:
Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.
def locked():
root = home()
root.mkdir(parents=True, exist_ok=True)
with open(root / ".lock", "w") as fh:
fcntl.flock(fh, fcntl.LOCK_EX)
try:
yield
Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.
existing = {e.get("id") for e in read_events()}
fresh = [e for e in new if e["id"] not in existing]
if fresh:
with open(events_path(), "a", encoding="utf-8") as fh:
for e in fresh:
fh.write(json.dumps(e, ensure_ascii=False) + "\n")
return fresh
Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.
s.bind(("127.0.0.1", 0))
port = s.getsockname()[1]
home().mkdir(parents=True, exist_ok=True)
log = open(home() / "onboard.log", "a")
subprocess.Popen(
[sys.executable, str(Path(__file__).resolve()), "onboard", "--port", str(port)],
stdout=log, stderr=log, stdin=subprocess.DEVNULL, start_new_session=True,
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
port = s.getsockname()[1]
home().mkdir(parents=True, exist_ok=True)
log = open(home() / "onboard.log", "a")
subprocess.Popen(
[sys.executable, str(Path(__file__).resolve()), "onboard", "--port", str(port)],
stdout=log, stderr=log, stdin=subprocess.DEVNULL, start_new_session=True,
)
Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.
port = s.getsockname()[1]
home().mkdir(parents=True, exist_ok=True)
log = open(home() / "onboard.log", "a")
subprocess.Popen(
[sys.executable, str(Path(__file__).resolve()), "onboard", "--port", str(port)],
stdout=log, stderr=log, stdin=subprocess.DEVNULL, start_new_session=True,
)
No suspicious patterns detected.