Is Terravision safe?
- Python shell/command execution
- Python dynamic code execution
- Dangerous MCP tool capability
What to do: Nothing here argues against installing it. Grant the capabilities it lists only if you expect the tool to need them.
Terravision is a PyPI package analyzed by SkillTotal's deterministic static scanner. The scan found no malicious indicators, though 1 risky construct is reported for review. It can: dynamic code execution, filesystem read, filesystem write, mcp tools detected, network egress and shell execution — capabilities are what the code can do, not a verdict on intent. Risk score 10/100 (low).
terravision 0.53.0
Automated static-analysis result. It can contain false positives and false negatives, and is not a claim about the intent of Terravision'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 (9)
The code turns strings into live code at runtime (eval / new Function / exec).
zone = str(eval(zone))
Why it matters: If those strings aren't fixed and trusted, they become a way to run arbitrary code.
Fix: Avoid evaluating dynamically constructed code; if unavoidable, ensure the input is a trusted constant and never derived from external data.
An MCP tool exposes a powerful capability (files, shell, network, browser, or credentials).
"name": "generate_interactive_html",
Why it matters: Wired into an agent, these grant it real access to your machine — confirm each is required.
Fix: Confirm each powerful tool is required and constrained; broad MCP tools (shell/filesystem/network) grant an agent significant host access.
An MCP server entry launches a command on your host.
"command": "uvx",
"command": "uvx",
"command": "uvx",
Why it matters: Trusting the manifest means running that binary — verify what it is and where it comes from.
Fix: Verify the launched command and its source before trusting this MCP server configuration.
The component can run operating-system commands or spawn processes.
subprocess.run(
[
"gvpr",
"-c",
"-q",
"-f",
str(path_to_script),
str(path_to_predot),
"-o",
str(path_to_postdot),
],subprocess.run(
["dot", "-Tpng", "-Gdpi=300", "-o", str(partial)],
input=source,
text=True,
check=True,
capture_output=True,
)import git # noqa: E402
from git import RemoteProgress # noqa: E402
result = subprocess.run(
["neato", "-Tdot"],
input="graph {}",
capture_output=True,
text=True,
timeout=30,
)result = subprocess.run(
[binary, "-v"], capture_output=True, text=True, check=True
)subprocess.run(["wslview", target], check=True)
result = subprocess.run(
["systemctl", "--user", "show-environment"],
env=env,
capture_output=True,
text=True,
timeout=5,
)subprocess.Popen(command, **popen_kwargs)
return subprocess.run(
["xdg-mime", "query", *args],
env=env,
stdin=subprocess.DEVNULL,
capture_output=True,
text=True,
timeout=10,
).stdout.strip()done = subprocess.run(
["open", str(path)],
stdin=subprocess.DEVNULL,
capture_output=True,
timeout=15,
)result = subprocess.run(
["dot", "-Txdot_json", "-o", json_file, dot_file],
capture_output=True,
text=True,
)result = subprocess.run(
init_cmd,
capture_output=not debug,
text=True,
stdin=subprocess.DEVNULL,
env=_tf_env(),
)result = subprocess.run(
[helpers.get_tf_binary(), "workspace", "select", "-or-create=True", workspace],
capture_output=not debug,
text=True,
stdin=subprocess.DEVNULL,
env=_tf_env(),
)result = subprocess.run(
plan_cmd,
capture_output=not debug,
text=True,
stdin=subprocess.DEVNULL,
env=_tf_env(),
)result = subprocess.run(
[helpers.get_tf_binary(), "show", "-json", tfplan_path],
stdout=f,
stderr=None if debug else subprocess.PIPE,
text=True,
stdin=subprocess.DEVNULL, …result = subprocess.run(
[helpers.get_tf_binary(), "graph"],
stdout=f,
stderr=None if debug else subprocess.PIPE,
text=True,
stdin=subprocess.DEVNULL,
env=_tf_env(), …result = subprocess.run(
["terragrunt", "--version"],
capture_output=True,
text=True,
timeout=30,
)result = subprocess.run(
init_cmd,
capture_output=not debug,
text=True,
cwd=workdir,
env=_tg_env(),
stdin=subprocess.DEVNULL,
)tf_result = subprocess.run(
[helpers.get_tf_binary(), "init", "-reconfigure", "-input=false"],
capture_output=not debug,
text=True,
cwd=cache_dir,
stdin=subprocess.DEVNULL,
)result = subprocess.run(
plan_cmd,
capture_output=not debug,
text=True,
cwd=workdir,
env=_tg_env(),
stdin=subprocess.DEVNULL,
)result = subprocess.run(
[helpers.get_tf_binary(), "show", "-json", tfplan_path],
stdout=f,
stderr=None if debug else subprocess.PIPE,
text=True,
cwd=cache_dir,
stdin=s …result = subprocess.run(
[helpers.get_tf_binary(), "graph"],
stdout=f,
stderr=None if debug else subprocess.PIPE,
text=True,
cwd=cache_dir,
stdin=subprocess.DEVNULL, …result = subprocess.run(
["neato", "-n2", "-Tjson0", dot_file_path],
capture_output=True,
text=True,
)subprocess.run(["npx", "--yes", MCPB_CLI, *args], check=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; avoid shell=True.
A server is bound to all network interfaces (0.0.0.0), not just your own machine.
stripped = sourceURL.split("::", 1)[1]prefix = sourceURL.split("::", 1)[0] + "::"Why it matters: Without authentication, other hosts on the network can reach it.
Fix: Bind to 127.0.0.1 for local-only use, or require authentication and restrict access if remote exposure is intended.
The component reads files from disk.
with open(path, "r", encoding="utf-8") as fh:
with open(path_to_postdot, "r", encoding="utf-8") as f:
with open(icon_path, "rb") as f:
with open(svg_path, "r", encoding="utf-8") as f:
with open(svg_path, "r", encoding="utf-8") as f:
with open(icon_path, "rb") as f:
with open(modules_json_path, "r") as f:
with open(full_filepath, "r") as fh:
with open(full_filepath, "r") as fh:
with open(cwd_ai_path, "r") as fh:
with open(annotate, "r") as file:
with open("/etc/os-release") as f:with open(source, "r") as fh:
_WSL_DETECTED = "microsoft" in Path("/proc/version").read_text().lower()return "data:image/svg+xml;base64," + base64.b64encode(path.read_bytes()).decode(
with open(template_path, "r", encoding="utf-8") as f:
with open(vendor_dir / "d3.min.js", "r", encoding="utf-8") as f:
text = candidate_path.read_text(
encoding="utf-8", errors="replace"
).strip()data = base64.b64encode(path.read_bytes()).decode("ascii")"text": resolved.read_text(encoding="utf-8"),
"blob": base64.b64encode(resolved.read_bytes()).decode("ascii"),xml = path.read_text(encoding="utf-8")
for entry in json.loads((patterns_dir / "index.json").read_text())
"graph": json.loads((patterns_dir / f"{pattern}.tvg.json").read_text()),rules = (_SKILL_DIR / "references" / "graph-format.md").read_text()
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.
os.remove(path_to_predot)
with open(svg_path, "w", encoding="utf-8") as f:
with open(temp_dot_path, "w", encoding="utf-8") as f:
os.remove(temp_dot_path)
os.remove(svg_path)
with open(drawio_output, "w", encoding="utf-8") as f:
os.remove(path_to_predot)
os.remove(path_to_postdot)
os.remove(path_to_predot)
os.remove(path_to_postdot)
os.unlink(tmp_path)
with open(sparse_checkout_file, "w") as f:
shutil.copytree(tf_cached_dir, target_dir)
shutil.copytree(codepath_module, temp_module_path)
shutil.rmtree(codepath, onerror=remove_readonly)
shutil.rmtree(codepath, ignore_errors=True)
with open(out_path, "w") as file:
with open(output_path, "w", encoding="utf-8") as f:
with open(output_path, "w", encoding="utf-8") as f:
with open(out_path, "w") as fh:
with open(target, "w", encoding="utf-8") as fh:
with open(target, "w", encoding="utf-8") as fh:
with open(graph_path, "w", encoding="utf-8") as fh:
page.write_text(
'<!doctype html><meta charset="utf-8">'
f"<title>{html.escape(path.name)} - draw.io</title>"
f"<script>location.replace({json.dumps(url)});</script>"
f'<p>Opening <a href="{html.escape(url)}" …with open(dest, "w") as f:
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 requests
resp = requests.get(download_url, stream=True, timeout=120)
response = requests.head(
download_endpoint,
headers=headers,
allow_redirects=False,
timeout=10,
)response = requests.get(
download_endpoint,
headers=headers,
allow_redirects=False,
timeout=10,
)disc_resp = requests.get(discovery_url, timeout=10)
r = requests.get(domain + gitaddress, headers=headers, timeout=10)
versions_resp = requests.get(versions_endpoint, headers=headers, timeout=10)
ver_resp = requests.get(
domain + gitaddress + "/" + resolved, headers=headers, timeout=10
)import requests
response = requests.get(f"{ollama_host}/api/tags", timeout=5)response = requests.get(url, timeout=5)
return requests.post(
url,
json=payload,
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {api_key}",
"Accept": "text/event-stream", …import urllib.request
with urllib.request.urlopen(url) as resp:
import urllib.request
with urllib.request.urlopen(SIDEBAR_AWS4_URL) as resp:
with urllib.request.urlopen(IMG_LIB_API) as resp:
with urllib.request.urlopen(cat_url) as resp2:
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.
An MCP tool surface (manifest or tool definitions) was found.
"mcpServers": {"mcpServers": {"mcpServers": {"tools": [
@mcp.tool()
@mcp.tool()
@mcp.tool()
@mcp.tool()
@mcp.tool(meta={"ui": {"visibility": ["app"]}})Why it matters: Just context — review which tools it offers and their permissions.
Fix: Review the declared MCP tools and their permissions.
How attackers abuse these capabilities
Interactive labs on the attack class behind the rules above. They show the technique, not anything found in Terravision.
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 →