Back to skill

Security audit

Design With Hopola

Security checks for vulnerabilities and agentic risk

Overview

Hopola is a coherent media-production workflow, but its upload helper can transmit any readable local file path to remote storage, which is broader than its stated asset-upload purpose.

Review before installing. This skill needs network access and an OPENCLOW_KEY for Hopola Gateway calls, which is expected, but its upload path should be treated carefully: only invoke upload on files you explicitly selected, avoid passing arbitrary local paths, and prefer a version that restricts uploads to session/generated asset directories, enforces the 20MB image limit, validates image content, and requires HTTPS for upload-related endpoints.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
scripts/maat_upload.py:281
Finding

Uploader Permits Transmission of Arbitrary Local Files

Content
View full analysis

Vulnerability Details

File Location: scripts/maat_upload.py:281-286, 333-337, 578-599, 670-678
Mirrored Location: Hopola-Skills/.trae/skills/Hopola/scripts/maat_upload.py:281-286, 333-337, 578-599, 670-678
Vulnerability Type: Missing local-path authorization and insufficient least-privilege enforcement
Risk Level: High

Vulnerable Code

python
def _normalize_local_path(path: str) -> str:
    cleaned = unicodedata.normalize("NFC", path).replace("\\", "/")
    collapsed = _collapse_path_slashes(cleaned)
    if not collapsed.startswith("/"):
        collapsed = os.path.abspath(collapsed)
    return os.path.normpath(collapsed)
python
def _normalize_session_reference(raw_value: str) -> str:
    candidates = _session_reference_candidates(raw_value)
    for candidate in candidates:
        if os.path.isfile(candidate):
            return candidate
python
def _multipart_body(fields: Dict[str, str], file_path: str, mime_type: str) -> Tuple[bytes, str]:
    boundary = f"----maat-{uuid.uuid4().hex}"
    boundary_bytes = boundary.encode("utf-8")
    with open(file_path, "rb") as f:
        file_bytes = f.read()
    file_name = os.path.basename(file_path)
    chunks: List[bytes] = []
    for name, value in fields.items():
        chunks.extend(
            [
                b"--" + boundary_bytes + b"\r\n",
                f'Content-Disposition: form-data; name="{name}"\r\n\r\n'.encode("utf-8"),
                str(value).encode("utf-8"),
                b"\r\n",
            ]
        )
python
def upload_local_file(file_path: str) -> str:
    normalized_file_path = _normalize_session_reference(file_path)
    suffix, mime_type = suffix_and_mime(normalized_file_path)
    raw_token = _request_upload_policy(suffix)
    token = _normalize_policy_token(raw_token)
    if _has_s3_token(token):
        cdn_url = _upload_via_s3(token, norm
...[truncated 2347 chars]
Remediation
View remediation

Remediation Suggestions

  1. Define an explicit set of permitted upload roots, preferably limited to directories dedicated to current-session attachments and generated assets.
  2. Resolve the candidate with os.path.realpath() or Path.resolve(strict=True) before authorization.
  3. Verify containment using os.path.commonpath() or Path.is_relative_to() against every approved root.
  4. Reject symlinks whose resolved targets leave an approved root.
  5. Reject arbitrary absolute paths and relative paths that resolve outside approved directories.
  6. Track attachment provenance and require the selected path to correspond to an actual user attachment or generated output.
  7. Require explicit user confirmation before uploading files outside the standard attachment directory.
  8. Open files only after authorization and, where supported, use no-follow semantics to reduce time-of-check/time-of-use and symlink risks.
  9. Apply the correction to both copies of maat_upload.py and add regression tests for absolute paths, traversal paths, symlink escapes, and valid session files.

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/maat_upload.py:385
Finding

Documented Upload Size and Image-Type Restrictions Are Not Enforced

Content
View full analysis

Vulnerability Details

File Location: scripts/maat_upload.py:385-397, 578-599
Related Configuration: config.template.json:174-199
Mirrored Location: Hopola-Skills/.trae/skills/Hopola/scripts/maat_upload.py:385-397, 578-599
Vulnerability Type: Missing file-size and content-type validation
Risk Level: Medium

Vulnerable Code

python
def suffix_and_mime(file_path: str) -> Tuple[str, str]:
    ext = file_path.rsplit(".", 1)[-1].lower() if "." in file_path else ""
    if ext in ("jpg", "jpeg"):
        return ext or "jpg", "image/jpeg"
    if ext == "png":
        return "png", "image/png"
    if ext == "webp":
        return "webp", "image/webp"
    if ext == "gif":
        return "gif", "image/gif"
    if ext == "svg":
        return "svg", "image/svg+xml"
    return "jpg", "image/jpeg"
python
def _multipart_body(fields: Dict[str, str], file_path: str, mime_type: str) -> Tuple[bytes, str]:
    boundary = f"----maat-{uuid.uuid4().hex}"
    boundary_bytes = boundary.encode("utf-8")
    with open(file_path, "rb") as f:
        file_bytes = f.read()
    file_name = os.path.basename(file_path)
    chunks: List[bytes] = []
    for name, value in fields.items():
        chunks.extend(
            [
                b"--" + boundary_bytes + b"\r\n",
                f'Content-Disposition: form-data; name="{name}"\r\n\r\n'.encode("utf-8"),
                str(value).encode("utf-8"),
                b"\r\n",
            ]
        )

The configuration declares a restriction that the executable uploader does not apply:

json
"request_log_default": "0",
"max_image_size_mb": 20,
"upload_payload_schema": {
  "files": "files",
  "metadata": "metadata"
}

Technical Analysis

File type is inferred only from the filename extension. Unknown extensions are silently treated as JPEG, without checking file signatures or decoding th ...[truncated 1670 chars]

Remediation
View remediation

Remediation Suggestions

  1. Enforce the configured maximum with os.path.getsize() before opening the file.
  2. Use a bounded or streaming multipart implementation rather than reading the complete file into memory.
  3. Reject unknown extensions instead of treating them as JPEG.
  4. Validate magic bytes and decode raster images with a maintained image library before upload.
  5. Confirm that decoded format and declared MIME type match.
  6. Apply image dimension and decompression-bomb limits in addition to the compressed file-size limit.
  7. Disable SVG uploads unless they are necessary. If required, sanitize SVG content and serve it with restrictive response headers.
  8. Apply equivalent server-side validation because client-side checks can be bypassed.
  9. Add tests for oversized files, extension spoofing, malformed images, polyglot files, decompression bombs, and active SVG content.

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/maat_upload.py:135
Finding

Credential Policy and Upload Endpoints Permit Plain HTTP

Content
View full analysis

Vulnerability Details

File Location: scripts/maat_upload.py:135-160, 346-356, 420-453, 525-537, 607-621
Mirrored Location: Hopola-Skills/.trae/skills/Hopola/scripts/maat_upload.py at the corresponding lines
Vulnerability Type: Sensitive network traffic allowed over unencrypted transport
Risk Level: Medium

Vulnerable Code

python
parsed = urllib.parse.urlsplit(value)
if parsed.scheme not in ("http", "https") or not parsed.netloc:
    raise MaatUploadError(
        f"Invalid upload policy endpoint: {value}",
        f"Configure a valid {TOKEN_API_ENV_NAME} value or use the default endpoint",
        MaatUploadError.VALIDATION_ERROR,
    )
host = _normalize_host(parsed.hostname or parsed.netloc)
allowed_hosts = _resolve_allowed_policy_hosts()
if host not in allowed_hosts:
    raise MaatUploadError(
        f"Upload policy endpoint host is not allowed: {host}",
        f"Review {TOKEN_API_ENV_NAME} or explicitly allow the host through {TOKEN_API_ALLOWED_HOSTS_ENV_NAME}",
        MaatUploadError.VALIDATION_ERROR,
    )

The messages above are translated for report readability; the executable condition is unchanged and explicitly permits both http and https.

python
def _normalize_delivery_url(url: str) -> str:
    value = _clean_str(url)
    parsed = urllib.parse.urlsplit(value)
    if parsed.scheme not in ("http", "https") or not parsed.netloc:
        raise MaatUploadError(
            f"Invalid upload delivery URL: {value}",
            "Retry later or review the upload credential configuration",
            MaatUploadError.VALIDATION_ERROR,
        )
    return _normalize_http_url(value)
python
req = urllib.request.Request(policy_url, method="GET")
try:
    with urllib.request.urlopen(req, timeout=30) as resp:
        code = resp.getcode() or 200
        raw = resp.read().decode("utf-8", errors="replace")
python
req =
...[truncated 2280 chars]
Remediation
View remediation

Remediation Suggestions

  1. Require parsed.scheme == "https" for policy, upload, access, data, and delivery URLs.
  2. Reject redirects that downgrade from HTTPS to HTTP.
  3. Validate the final redirect destination after every redirect, including its scheme and hostname.
  4. Maintain a separate allowlist for storage upload hosts rather than trusting any host returned by the policy response.
  5. Restrict policy-host configuration to trusted administrators and avoid expanding the allowlist from untrusted workflow input.
  6. Use standard certificate and hostname verification without disabling TLS validation.
  7. Ensure temporary credentials have short lifetimes, narrow object-key prefixes, minimal operations, and no read or list permissions unless required.
  8. Apply the HTTPS checks to both copies of the uploader and add regression tests for HTTP URLs, HTTPS-to-HTTP redirects, and unauthorized storage hosts.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (110)

Concealed Executable Artifact

High
Category
Supply Chain
Confidence
100% confidence
Finding

An executable nested in a document or hidden/disguised artifact can evade ordinary extension-based review while still being available to the skill at runtime.

Content

No source excerpt is available for this finding.

Concealed Executable Artifact

High
Category
Supply Chain
Confidence
100% confidence
Finding

An executable nested in a document or hidden/disguised artifact can evade ordinary extension-based review while still being available to the skill at runtime.

Content

No source excerpt is available for this finding.

Concealed Executable Artifact

High
Category
Supply Chain
Confidence
100% confidence
Finding

An executable nested in a document or hidden/disguised artifact can evade ordinary extension-based review while still being available to the skill at runtime.

Content

No source excerpt is available for this finding.

Concealed Executable Artifact

High
Category
Supply Chain
Confidence
100% confidence
Finding

An executable nested in a document or hidden/disguised artifact can evade ordinary extension-based review while still being available to the skill at runtime.

Content

No source excerpt is available for this finding.

Concealed Executable Artifact

High
Category
Supply Chain
Confidence
100% confidence
Finding

An executable nested in a document or hidden/disguised artifact can evade ordinary extension-based review while still being available to the skill at runtime.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description suggests a production pipeline involving web research, media generation, uploads, and report creation. The supplied code does none of those things. Its sole purpose is to build a release ZIP from local files while excluding certain directories and file types. This is a materially different primary purpose and involves local filesystem packaging behavior that is not represented in the description.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description says the skill runs a full multimodal production and delivery pipeline. However, the actual code does not perform any of those operational tasks. It only validates configuration structure and expected constants in a JSON config file, such as preferred tool names, required schema lists, and supported stage names. This is a materially different primary purpose, so the description does not accurately represent the supplied code chunk.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code chunk is narrowly focused on uploading an existing local or session file. It requests an upload policy from a remote endpoint, constructs either S3-style signed form uploads or a legacy token-based upload, performs the upload, checks accessibility of the returned URL, and prints the resulting URL. There is nothing in the code related to web research, orchestration of generation models, multimodal asset production, or Markdown report creation. While 'upload' is one small element mentioned in the description, the declared purpose describes a broad end-to-end workflow that is not represented here. Additionally, the code performs network calls to retrieve upload credentials and upload content, which is a meaningful operational capability not reflected by the declared empty permissions. Therefore the description does not accurately represent the actual behavior of this code chunk.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Yes, this is a mismatch. The declared description claims a broad, end-to-end multimodal production and delivery workflow, including research, media generation, reporting, and uploading. The actual code chunk only accepts a local file path and calls another script to upload that file. Uploading is one small subset of the declared behavior, but the primary purpose of the supplied code is not the full workflow described. No evidence in this chunk supports the research, generation, orchestration, or Markdown reporting capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code does not perform web research, media generation orchestration, uploading, or report production. Instead, it validates that a skill package is ready for release by inspecting local files and contents for required artifacts, security issues, and documentation guardrails. This is a materially different primary purpose from the declared multimodal production workflow, so the description does not accurately represent the code's actual behavior.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 32)May include surrounding context.

md
- `SKILL.md`: Main orchestration and strategy

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/examples.md (reported line 477)May include surrounding context.

1)发现工具

bash
curl -s "https://hopola.ai/api/gateway/mcp/tools" \
  -H "X-OpenClaw-Key: $OPENCLOW_KEY"

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 33)May include surrounding context.

md
"assets/logo.svg",
    "assets/cover.svg",
    "assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 19)May include surrounding context.

python
"assets/logo.svg",
    "assets/cover.svg",
    "assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 34)May include surrounding context.

md
"assets/cover.svg",
    "assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 20)May include surrounding context.

python
"assets/cover.svg",
    "assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 35)May include surrounding context.

md
"assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 21)May include surrounding context.

python
"assets/flow.svg",
    "subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 36)May include surrounding context.

md
"subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 22)May include surrounding context.

python
"subskills/search/SKILL.md",
    "subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 37)May include surrounding context.

md
"subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",
    "subskills/upload/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 23)May include surrounding context.

python
"subskills/generate-image/SKILL.md",
    "subskills/generate-video/SKILL.md",
    "subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",
    "subskills/upload/SKILL.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 39)May include surrounding context.

md
"subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",
    "subskills/upload/SKILL.md",
    "subskills/report/SKILL.md",
    "playbooks/design-intake.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/scripts/validate_release.py (reported line 25)May include surrounding context.

python
"subskills/generate-3d/SKILL.md",
    "subskills/logo-design/SKILL.md",
    "subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",
    "subskills/upload/SKILL.md",
    "subskills/report/SKILL.md",
    "playbooks/design-intake.md",

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · Hopola-Skills/.trae/skills/Hopola/SKILL.md (reported line 41)May include surrounding context.

md
"subskills/product-image/SKILL.md",
    "subskills/cutout/SKILL.md",
    "subskills/upload/SKILL.md",
    "subskills/report/SKILL.md",
    "playbooks/design-intake.md",
    "scripts/check_tools_mapping.py",
    "scripts/build_release_zip.py",