Back to skill

Security audit

marketing

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly aligned with Instagram content generation, but it sends the API key and product page data to an API hostname that differs from the main documentation.

Review before installing or using. Only use a narrowly scoped SKILLBOSS_API_KEY, confirm whether api.heybossai.com is an authorized SkillBoss endpoint, avoid private or sensitive product URLs, and treat extracted product claims as untrusted until checked against the source page.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/extract_product.py:28
Finding

API Credential Sent to a Destination Inconsistent with User-Facing Documentation

Content
View full analysis
dict: r = requests.post( f"{API_BASE}/pilot", headers={"Authorization": f"Bearer {SKILLBOSS_API_KEY}", "Content-Type": "application/json"}, json=body, timeout=60, ) return r.json() ``` ### Technical Analysis The script reads `SKILLBOSS_API_KEY` from the process environment and transmits it as an HTTP bearer credential on every `pilot()` request. Although the request uses HTTPS, the receiving hostname is inconsistent with the endpoint disclosed in the user-facing Skill instructions. The script's own docstring mentions `api.heybossai.com`, but users following `SKILL.md` are told that their credential will be used with `api.skillbossai.com`. This inconsistency prevents informed verification of the credential recipient and creates a trust-boundary violation. Remote API access is relevant to the declared product-extraction functionality. However, sending a credential to an endpoint different from the documented endpoint exceeds the network authority that users can reasonably infer from the primary Skill instructions. The available evidence does not establish whether the two domains have common ownership, so malicious intent cannot be asserted. ### Attack Path 1. A user installs ...[truncated 1260 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/extract_product.py:50
Finding

Indirect Prompt Injection Through Untrusted Scraped Page Content

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Tainted flow: 'SKILLBOSS_API_KEY' from os.environ (line 27, credential/environment) → requests.post (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/extract_product.py (reported line 32)May include surrounding context.

python
def pilot(body: dict) -> dict:
    r = requests.post(
        f"{API_BASE}/pilot",
        headers={"Authorization": f"Bearer {SKILLBOSS_API_KEY}", "Content-Type": "application/json"},
        json=body,

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The declared behavior says the skill generates Instagram marketing content, but the documented implementation also performs external scraping and LLM-based extraction through SkillBoss API Hub, which is materially different and more sensitive than pure content generation. This mismatch can mislead users and reviewers into approving a skill without realizing it transmits URLs and scraped content externally, enabling unexpected data handling and broader capability use than advertised.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill requires an API key and explicitly instructs use of a script that performs external network access, but it does not declare any tool scope or permissions boundary. This creates a trust and review gap: operators and users cannot easily see that the skill can access environment secrets and send data off-platform, increasing the risk of unintended secret use or network exfiltration.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The skill contains an explicit external endpoint and directs use of it for scraping and LLM extraction, which means user-provided inputs and retrieved page data are sent outside the local trust boundary. External transmission is not inherently malicious, but it is dangerous when not tightly scoped, disclosed, and governed because it creates data leakage, third-party dependency, and compliance risks.

Content

Scanner excerpt · SKILL.md (reported line 79)May include surrounding context.

SKILLBOSS_API_KEY=your_key python3 scripts/extract_product.py

text

The script uses **SkillBoss API Hub** (`https://api.skillbossai.com/v1/pilot`) to:
1. Scrape the product page via the `scraping` capability
2. Extract structured product data via the `chat` capability (auto-routed LLM)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill instructs users to send product URLs and scraped page content to a third-party API without any privacy notice, consent flow, or explanation of retention and downstream processing. Even if product pages are often public, transmitted content may include tracking parameters, merchant-specific data, or user-supplied URLs that reveal business intent, making silent external sharing a real privacy and compliance concern.

Content

No source excerpt is available for this finding.

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 · scripts/extract_product.py (reported line 8)May include surrounding context.

python
Extracts product information from e-commerce URLs for Instagram content generation.
Supports: Amazon, Shopify, Taobao, JD, and generic e-commerce sites.

Uses SkillBoss API Hub (https://api.heybossai.com/v1/pilot) for web scraping
and AI-powered structured data extraction.

Usage:

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 · scripts/extract_product.py (reported line 28)May include surrounding context.

python
Extracts product information from e-commerce URLs for Instagram content generation.
Supports: Amazon, Shopify, Taobao, JD, and generic e-commerce sites.

Uses SkillBoss API Hub (https://api.heybossai.com/v1/pilot) for web scraping
and AI-powered structured data extraction.

Usage:

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% 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 · scripts/extract_product.py (reported line 32)May include surrounding context.

python
def pilot(body: dict) -> dict:
    r = requests.post(
        f"{API_BASE}/pilot",
        headers={"Authorization": f"Bearer {SKILLBOSS_API_KEY}", "Content-Type": "application/json"},
        json=body,

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script sends the user-supplied product URL and up to 3000 characters of scraped page content to a third-party API for extraction, but it provides no explicit disclosure, consent flow, or data handling warning. If users provide private, non-public, or sensitive product pages, this can cause unintended data sharing outside the local environment.

Content

No source excerpt is available for this finding.

Ssd 1

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

Untrusted page text is interpolated directly into the LLM prompt, so a malicious product page can include prompt-injection content that manipulates the model's extraction behavior, produces malformed JSON, or returns attacker-chosen fields. In this skill context, the model output is then trusted and surfaced as structured marketing input, making content poisoning and downstream misinformation plausible.

Content

No source excerpt is available for this finding.