Is Watermarks Remover safe?
- Python shell/command execution
- Node.js shell/command execution
- Python filesystem read
What to do: Nothing here argues against installing it. Grant the capabilities it lists only if you expect the tool to need them.
Watermarks Remover is a directory analyzed by SkillTotal's deterministic static scanner. The scan found no malicious indicators. It can: filesystem read, filesystem write, network egress and shell execution — capabilities are what the code can do, not a verdict on intent. Risk score 0/100 (low).
repo
Automated static-analysis result. It can contain false positives and false negatives, and is not a claim about the intent of Watermarks Remover's authors. Report a false positive.
Behavioral traits
How this component maps to the CSA agentic threat model. Descriptive — it never affects the risk score.
Findings (5)
The component can run operating-system commands or spawn processes.
const { spawn, execFile } = require('child_process');// that failed to spawn (ENOENT) cannot let its later `close` event terminate
const child = spawn(cmd, args, { stdio: 'inherit', windowsHide: true });Why it matters: Powerful and often legitimate — confirm the commands aren't built from untrusted input.
Fix: Confirm the command and its arguments are fully controlled and not derived from untrusted input; prefer execFile with an argument array.
The component can run operating-system commands or spawn processes.
r = subprocess.run(
[git, "-C", str(upstream), "rev-parse", "HEAD"],
capture_output=True,
text=True,
timeout=10,
check=False,
creationflags=subprocess_creationflags, …r = subprocess.run(
[git, "-C", str(repo_root), "rev-parse", "HEAD"],
capture_output=True,
text=True,
timeout=10,
check=False,
creationflags=subprocess_creationflags, …return subprocess.run(
cmd,
capture_output=True,
text=True,
timeout=timeout,
check=False,
creationflags=subprocess_creationflags,
)self._proc = subprocess.Popen(
cmd,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
text=True,
bufsize=1,
creationflags=subprocess_cr …r = subprocess.run(
[
ffprobe,
"-v",
"error",
"-select_streams",
"a:0",
"-show_entries",
"stream=sample_rate", …r = subprocess.run(
[
ffprobe,
"-v",
"error",
"-show_entries",
"stream=codec_type",
"-of",
"csv=p=0", …proc = subprocess.run(
[sys.executable, str(CLEAN_FILE_PY), str(path), "--in-place", "--json"],
capture_output=True,
text=True,
check=False,
creationflags=subprocess_creationflags,
)r = subprocess.run(
cmd,
capture_output=True,
text=True,
timeout=timeout,
preexec_fn=subprocess_preexec_fn,
check=False,
creationflags=subprocess_creationfl …r = subprocess.run(
[
ffprobe,
"-v",
"error",
"-select_streams",
"v:0",
"-show_entries",
"stream=r_frame_rate", …r = subprocess.run(
[
ffprobe,
"-v",
"error",
"-select_streams",
"v:0",
"-show_entries",
"frame=pts_time", …r = subprocess.run(
[qpdf, "--linearize", "--", safe_arg(str(dest)), safe_arg(str(tmp))],
capture_output=True,
text=True,
timeout=(_QPDF_TIMEOUT if deadline is None else deadline.timeout(_QPDF …r = subprocess.run(
[exiftool, "-all=", "-overwrite_original", safe_arg(str(dest))],
capture_output=True,
text=True,
timeout=(
_EXIFTOOL_TIMEOUT if deadline is None else deadli …r = subprocess.run(
[
gs,
"-dBATCH",
"-dNOPAUSE",
"-dQUIET",
"-dSAFER",
"-sDEVICE=pdfwrite",
"-dPDFSETTINGS=/prepres …r = subprocess.run(
[qpdf, "--json", "--json-key=attachments", "--", safe_arg(str(path))],
capture_output=True,
text=True,
timeout=(
_QPDF_ATTACH_TIMEOUT if deadline is None el …p = subprocess.Popen(
[qpdf, f"--show-attachment={key}", "--", safe_arg(str(path))],
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
# Same preexec_fn resource-limit guard u …r = subprocess.run(
[qpdf, *cmd],
capture_output=True,
text=True,
timeout=(
_QPDF_ATTACH_TIMEOUT if deadline is None else deadline.timeout(_QPDF_ATTACH_TIMEOUT)
), …proc = subprocess.run(
[sys.executable, str(CLEAN_FILE_PY), str(path), "-o", str(temp_path), "--json"],
capture_output=True,
text=True,
check=False,
creationflags=subprocess_creati …r = subprocess.run(
[c2patool, safe_arg(str(path))],
capture_output=True,
text=True,
timeout=30,
preexec_fn=subprocess_preexec_fn,
check=False, …r = subprocess.run(
[exiftool, "-G1", "-a", "-s", safe_arg(str(path))],
capture_output=True,
text=True,
timeout=30,
preexec_fn=subprocess_preexec_fn, …r = subprocess.run(
cmd,
capture_output=True,
text=True,
timeout=180,
preexec_fn=subprocess_preexec_fn,
check=False,
creationflags=subprocess_creationflags, …r = subprocess.run(
cmd,
capture_output=True,
text=True,
timeout=timeout,
preexec_fn=ctrlregen_subprocess_preexec_fn,
check=False,
creationflags=subprocess_ …r = subprocess.run(
cmd,
capture_output=True,
text=True,
timeout=timeout,
preexec_fn=ctrlregen_subprocess_preexec_fn,
check=False,
creationflags=subprocess_ …subprocess.run(
[
exiftool,
"-all=",
"-overwrite_original",
safe_arg(str(dest)),
],
capture_output=True, …r = subprocess.run(
[gs, "--version"],
capture_output=True,
text=True,
timeout=10,
check=False,
preexec_fn=subprocess_preexec_fn,
creationflags=subprocess_c …r = subprocess.run(
[path, _VERSION_FLAG.get(cmd, "--version")],
capture_output=True,
text=True,
timeout=10,
preexec_fn=subprocess_preexec_fn,
check=False, …Why it matters: Powerful and often legitimate — confirm the commands aren't built from untrusted input.
Fix: Confirm the command and its arguments are fully controlled and not derived from untrusted input; avoid shell=True.
The component reads files from disk.
fields = parse_frontmatter(skill_md.read_text(encoding="utf-8"))
bundle.writestr(info, (source / rel).read_bytes())
text = path.read_text(encoding="utf-8", errors="surrogateescape")
text = path.read_text(encoding="utf-8", errors="surrogateescape")
data = path.read_bytes()
data = path.read_bytes()
watermarked = wm_path.read_text(encoding="utf-8", errors="surrogateescape")
unwatermarked = plain_path.read_text(encoding="utf-8", errors="surrogateescape")
out_text = Path(out_path).read_text(encoding="utf-8", errors="surrogateescape")
data = f.read_text(encoding="utf-8", errors="surrogateescape").strip()
prompt = prompt_path.read_text(encoding="utf-8", errors="surrogateescape")
container_fmt = detect_container_format(args.path, args.path.read_bytes())
raw = args.path.read_bytes() if kind == "text" else None
(src.read_bytes() != dest.read_bytes())
(src.read_bytes() != dest.read_bytes())
(src.read_bytes() != dest.read_bytes())
return p_in.read_bytes() != p_out.read_bytes()
f.write(src.read_bytes())
safe_write_bytes(dest, tmp.read_bytes())
data = tmp.read_bytes()
data = in_path.read_bytes()
safe_write_bytes(dest, tmp.read_bytes())
return dest.read_bytes(), result.get("actions", [])return dest.read_bytes(), result.get("actions", [])return dest.read_bytes(), result.get("actions", [])Why it matters: Usually legitimate, but worth confirming it can't be steered into reading sensitive files.
Fix: Confirm which files are read and that paths cannot be influenced by untrusted input to reach sensitive locations.
The component writes or deletes files on disk.
shutil.copytree(
source,
staged_skill,
ignore=shutil.ignore_patterns(*EXCLUDED_NAMES, "*.pyc", "*.pyo"),
)shutil.rmtree(staging_root, ignore_errors=True)
shutil.rmtree(staging_root, ignore_errors=True)
tmp.write_bytes(data)
safe_write_bytes(dest, cleaned)
os.unlink(path)
prompt_path.write_text(prompt, encoding="utf-8", errors="surrogateescape")
in_path.write_text(entry["input"], encoding="utf-8", errors="surrogateescape")
out_path.write_text(entry["output"], encoding="utf-8", errors="surrogateescape")
(out_dir / "report.md").write_text(report, encoding="utf-8")
(out_dir / "results.json").write_text(json.dumps(payload, indent=2), encoding="utf-8")
(out_dir / "results.csv").write_text("\n".join(csv_lines) + "\n", encoding="utf-8")safe_write_bytes(output, data)
safe_write_text(dest, cleaned)
list_path.write_text("\n".join(lines) + "\n", encoding="utf-8")os.unlink(tmp_out)
safe_write_text(path, text)
os.unlink(tmp_name)
safe_write_bytes(path, text.encode("utf-8", errors="surrogateescape"))os.unlink(bak)
safe_write_bytes(dest, tmp.read_bytes())
safe_write_bytes(clean_path, cleaned_bytes)
safe_write_bytes(dest, tmp.read_bytes())
safe_write_bytes(dest, blanked if n else source)
safe_write_bytes(dest, data)
Why it matters: Usually legitimate, but worth confirming the paths can't be controlled by untrusted input.
Fix: Confirm which files are written/deleted and that paths cannot be influenced by untrusted input.
The component makes outbound network requests.
import http.client
import urllib.request
path = urllib.parse.urlparse(url).path.lower()
url_filename = Path(urllib.parse.urlsplit(url).path).name or "input"
parsed = urllib.parse.urlsplit(url)
conn = http.client.HTTPSConnection(
host,
port,
timeout=timeout,
context=context,
)conn = http.client.HTTPConnection(host, port, timeout=timeout)
parsed = urllib.parse.urlsplit(url)
current_url = urllib.parse.urljoin(current_url, location)
import urllib.request
req = urllib.request.Request(url, method=method) # noqa: S310 - scheme checked above
with urllib.request.urlopen(req, timeout=30.0) as resp: # noqa: S310
import urllib.request
raise urllib.error.HTTPError(req.full_url, code, msg, headers, fp)
req = urllib.request.Request( # noqa: S310
base_url.rstrip("/") + "/score",
data=body,
headers=headers,
method="POST",
)with urllib.request.urlopen(req, timeout=timeout) as resp: # noqa: S310
opener = urllib.request.build_opener(_NoRedirect())
import urllib.request
raise urllib.error.HTTPError(req.full_url, code, msg, headers, fp)
req = urllib.request.Request( # noqa: S310
url,
data=body,
headers={"Content-Type": "application/json", **headers},
method="POST",
)opener = urllib.request.build_opener(_NoRedirect())
import urllib.request
raise urllib.error.HTTPError(req.full_url, code, msg, headers, fp)
return urllib.request.build_opener(_NoRedirect())
req = urllib.request.Request( # noqa: S310
base_url.rstrip("/") + "/watermark",
data=payload_bytes,
headers=headers,
method="POST",
)Why it matters: Usually legitimate, but confirm the destinations are expected and no sensitive data leaves.
Fix: Confirm the destination hosts are expected and that no sensitive data is sent off-host.
How attackers abuse these capabilities
Interactive labs on the attack class behind the rules above. They show the technique, not anything found in Watermarks Remover.
Check your own component
Run the same evidence-backed scan on any MCP server, agent skill, or package.
Scan your own componentHow we determine this: deterministic static analysis (regex + AST), evidence-anchored, no code execution. Methodology →