Back to skill

Security audit

NoteGPT AI Creative Slides Skill

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report missing disclosures about sending content to external image providers and several lower-confidence code, permission-scope, and dependency concerns; the evidence is incomplete.

Consider the suspicious verdict before installing, especially if you would use confidential source material or images: the reported workflow may send them to an external provider without an explicit disclosure or authorization check. Skillspector scanned all listed components but marked its analysis partial because some local references were unresolved; the separate AIG scan completed with no reported findings, which does not establish that the skill is safe.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
Findings (45)

Tainted flow: 'req' from os.getenv (line 192, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/image_ppt_runtime.py (reported line 198)May include surrounding context.

python
method="GET",
    )
    try:
        with urllib.request.urlopen(req, timeout=timeout) as resp:
            print(f"models endpoint: HTTP {resp.status}")
            return 200 <= resp.status < 300
    except urllib.error.HTTPError as exc:

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/image-model-configuration.md (reported line 79)May include surrounding context.

The config is written to:

text
~/.image-ppt-skill/.env

The file is created with mode 0600. It is shared by Claude Code, OpenClaw, Hermes Agent, and other local agents.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/image_gen.py (reported line 69)May include surrounding context.

python
def _runtime_env_path() -> Path:
    return _runtime_home() / ".env"


def _load_runtime_env() -> None:

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/image_ppt_runtime.py (reported line 45)May include surrounding context.

python
def _env_path(home: Path) -> Path:
    return home / ".env"


def _requirements_path() -> Path:

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/image_gen.py (reported line 191)May include surrounding context.

python
_die(f"Prompt file not found: {path}")
        return path.read_text(encoding="utf-8").strip()
    if prompt:
        return prompt.strip()
    _die("Missing prompt. Use --prompt or --prompt-file.")
    return ""  # unreachable

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/image_gen.py (reported line 380)May include surrounding context.

python
def _augment_prompt_fields(augment: bool, prompt: str, fields: Dict[str, Optional[str]]) -> str:
    if not augment:
        return prompt

    sections: List[str] = []
    if fields.get("use_case"):

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/image_ppt_runtime.py (reported line 262)May include surrounding context.

python
bootstrap.add_argument("--upgrade", action="store_true")
    bootstrap.set_defaults(func=_bootstrap)

    config = subparsers.add_parser("config", help="Write or update shared .env")
    config.add_argument("--api-key")
    config.add_argument("--base-url")
    config.add_argument("--clear-base-url", action="store_true")

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

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.

Content

Scanner excerpt · SKILL.md (reported line 18)May include surrounding context.

md
## Hard Constraints

- Assemble and deliver `.pptx` by default; do not ask users to choose between slide images and PPTX. Images are intermediate assets or supplementary deliverables unless the user explicitly requests images only.
- Read the relevant `Reference Map` files before each phase. This file is the orchestration contract; detailed rules live in `docs/` and per-slide guidance in `prompts/`.
- For every script used in this workflow, learn usage from the relevant documentation first. If it does not answer the concrete question, run the script's `--help` (and subcommand `--help` when applicable). Read source only after both documentation and help are unavailable, fail, or leave that same question unresolved; then inspect only the relevant sections. An execution error alone does not bypass this order.
- First confirm visual style, audience, scenario, tone, and slide count together, reusing explicitly specified or clearly inferable values without confirmation. Require outline and

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

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.

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

md
from the relevant documentation first. If it does not answer the concrete question, run the script's `--help` (and subcommand `--help` when applicable). Read source only after both documentation and help are unavailable, fail, or leave that same question unresolved; then inspect only the relevant sections. An execution error alone does not bypass this order.
- First confirm visual style, audience, scenario, tone, and slide count together, reusing explicitly specified or clearly inferable values without confirmation. Require outline and sample confirmation afterward. Skip sample confirmation also when inspection verifies consistency with a user-provided style reference image. Otherwise waive confirmation only on explicit user request, following `docs/workflow-gates-and-progress.md`; never infer a waiver from silence or a general creation request.
- Confirmation means asking about unresolved settings or the displayed outline/sample and waiting for the user's actual selection or affirmati

Image-generation workflow lacks a disclosure that source content and reference images may be sent to an external provider.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow can transmit source-derived slide content and reference images to an image-generation provider without explicitly explaining that data transfer. Reports, papers, and internal notes can contain confidential information, and outline or visual approval does not establish permission to disclose it to an external service.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

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.

Content

Scanner excerpt · docs/backend-selection.md (reported line 17)May include surrounding context.

md
ed parameters.
- Select CLI/API fallback directly without a separate approval gate, only when no built-in image generation tool is available.
- Do not recommend CLI/API fallback merely because it provides direct `--out` file paths, easier local file management, local config reuse, batch generation convenience, or simpler automation.
- Before generating the first image, tell the user which tool availability you checked, which backend you plan to use, why fallback is or is not needed, and proceed without asking for backend confirmation. Do not treat being in a specific agent environment as proof that the built-in image tool is available.
- CLI/API fallback loads `~/.image-ppt-skill/.env` automatically. Run the CLI normally; do not manually parse `.env` or ask for configuration before an error.
- Ask for `OPENAI_API_KEY` configuration only after you have intentionally selected CLI/API fallback and that fallback reports missing config, after authentication/base URL/model errors, or when th

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

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.

Content

Scanner excerpt · docs/backend-selection.md (reported line 27)May include surrounding context.

md
## Report The Selection

Briefly state the selected backend and the reason when helpful, then continue. Do not ask the user to approve the backend. Outline, visual-style, and sample-slide approval requirements remain unchanged.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

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.

Content

Scanner excerpt · docs/cli-api-fallback.md (reported line 15)May include surrounding context.

python3 {skill_root}/scripts/image_ppt_runtime.py bootstrap

text

This is an internal setup step for the skill. Do not ask the user to run it unless dependency installation fails and user approval or troubleshooting is required.

The fallback CLI loads `~/.image-ppt-skill/.env` automatically for `OPENAI_API_KEY`, `OPENAI_BASE_URL`, and `IMAGE_PPT_IMAGE_MODEL`. Do not manually parse `.env`. Read `image-model-configuration.md` together with this file when no built-in image generation tool is available. Reading the configuration guide does not require inspecting secrets or asking for configuration unless needed.

API generation and editing instructions omit a warning about uploading prompt content and slide images.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The documented API commands send prompt text and, for image-input operations, slide or source images to the configured provider. Without a disclosure or sensitivity check, an agent can upload proprietary report content or private imagery while treating generation as routine local processing.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

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.

Content

Scanner excerpt · docs/image-model-configuration.md (reported line 22)May include surrounding context.

md
- The agent uses a third-party API or OpenAI-compatible proxy for image generation.
- The skill is being used from Claude Code, OpenClaw, Hermes Agent, or another agent without a built-in image tool.

If a built-in image tool is available, do not ask the user to configure `gpt-image-2.5-flare`.

## Required And Optional Values

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · docs/image-model-configuration.md (reported line 70)May include surrounding context.

bash
python3 {skill_root}/scripts/image_ppt_runtime.py config \
  --api-key "your-atlascloud-api-key" \
  --base-url "https://api.atlascloud.ai/api/v1/model" \
  --model openai/gpt-image-2

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

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.

Content

Scanner excerpt · docs/outline-style-and-sample.md (reported line 17)May include surrounding context.

md
- Tone: for example, neutral, friendly, professional, educational, technical, or academic.
- Slide count: total slides, including the cover and closing; suggest a suitable count when unspecified and confirm it with the other unresolved settings. Keep the outline and final deck at that count unless the user approves a change.

Reuse explicitly specified values and settings clearly inferable from the user's request without asking again. Briefly state inferred settings without claiming user approval; ask only about missing or ambiguous values in one grouped question, preferably using an available built-in question tool, suggesting a suitable combination rather than separate confirmation rounds. Wait for the user's choices unless confirmation is explicitly waived; when waived, choose suitable defaults and state the assumptions. If all settings are explicit or clearly inferable, briefly restate them and continue without waiting. Record the resolved settings in `outline.md` once the project

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

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.

Content

Scanner excerpt · docs/outline-style-and-sample.md (reported line 85)May include surrounding context.

md
tle and all text intended to appear in the PPT, including subtitles, body text, labels, captions, chart/table text, and footers where applicable. Keep the presentation concise through formatting, not by summarizing or omitting slide text. Speaker notes and internal layout or generation instructions are excluded.

Show the outline and request confirmation, preferably using an available built-in question tool, then wait for approval before image generation, unless the user explicitly asked you to skip confirmation. If any slide lists required source images, explicitly ask the user to verify that each image is assigned to the correct slide and role before generation. If the user requests changes, update `outline.md` and ask for confirmation again.

Unless the user explicitly waived outline confirmation, stop after writing the outline draft. At this point, report the `outline.md` path, slide count, required source images and their slide mapping, and that no slide images or PPTX have been g

Sample-generation instructions do not disclose external transmission of outline content and references.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

Generating the sample can disclose approved outline text and reference imagery before the user sees the result. The detailed visual approval process does not explain the external transfer or establish that confidential source material may be sent to that provider.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

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.

Content

Scanner excerpt · docs/outline-style-and-sample.md (reported line 130)May include surrounding context.

md
e exception applies.
- Compare the sample visually with any user-provided style reference image: check palette, typography hierarchy, density and whitespace, texture, and illustration/icon treatment. Match the visual identity while allowing a layout appropriate to the sample content. Also perform normal text, content, and required-asset QA. If the style matches and QA passes, show the sample, briefly report the match, record the reference path and comparison result in `outline.md`, and continue without asking for sample confirmation. A reference image existing or appearing in the prompt is insufficient evidence. If the match is uncertain or inconsistent, ask the user to confirm or adjust the visual style, typography, layout density, and text quality.

When sample confirmation is required, prefer an available built-in question tool after displaying the image. Do not generate the full deck until the user approves the sample slide or explicitly waives sample confirmation, or the verified-

API-based slide repair lacks a disclosure that the existing slide is uploaded.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The repair workflow directs the agent to submit an existing slide to an image API without first checking whether external processing is authorized. Slides derived from private reports may contain sensitive information, so a seemingly local repair can cross a confidentiality boundary.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

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.

Content

Scanner excerpt · docs/project-assembly-and-reporting.md (reported line 108)May include surrounding context.

python3 {skill_root}/scripts/image_ppt_runtime.py bootstrap

text

This is an internal setup step for the skill. Do not ask the user to run these commands unless dependency installation fails and user approval or troubleshooting is required.

Run:

The reusable prompt template hardcodes Chinese without a user language choice.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The generic slide-prompt template sets the output language to Chinese rather than using the presentation's selected language. Unlike the later conditional guidance for Chinese decks, this field is unconditional and can impose Chinese on decks requested in another language.

Content

No source excerpt is available for this finding.

Batch generation attaches source assets without requiring a privacy disclosure.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow requires attaching source assets and style references to image-generation calls without an explicit external-processing authorization check. Local documents or figures may contain confidential information, and approving slide content does not necessarily authorize uploading those assets to the selected service.

Content

No source excerpt is available for this finding.