Back to skill

Security audit

admapix

Security checks for vulnerabilities and agentic risk

Overview

This AdMapix skill appears to be a read-only external data client, but it needs review because its broad triggers and command/API fallback are under-scoped for the authority it may use.

Install only if you intend to send AdMapix ad-intelligence queries and filters to AdMapix using an API key. Prefer a scoped/protocol tool over direct curl execution, use a limited dedicated API key, and avoid invoking it for unrelated app or data questions until the trigger scope and capability documentation are tightened.

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
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

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 · SKILL.md (reported line 151)May include surrounding context.

md
For the creative `search` endpoint, `page_size` is capped at **10** (clamp any larger request down to 10; use `page` for more). Other list endpoints use their own documented ranges.

## Output Rules

Return the API response as **raw structured JSON** — keep the API field names; do not rename, drop, summarize, rank, or editorialize. The calling agent composes and analyzes.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Broad trigger keywords like 'apps', 'data', and 'layer' can cause the skill to activate in many unrelated conversations. Over-triggering can route unrelated user requests into an external-data skill, leading to unnecessary third-party requests and increased exposure of user prompts or query terms.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The main description explicitly says the skill only fetches raw JSON and does not analyze, summarize, recommend, or generate pages. Later documentation claims capabilities such as automated data analysis and report generation, which directly contradict the earlier stated intent and the documented output rules.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The hardcoded external service endpoint means the skill is designed to move data to a third-party API. While this is core functionality, it is still security-relevant because user requests, filters, or business queries may leave the local trust boundary.

Content

Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.

bash
admapix_auth_header="X-API-Key: ${ADMAPIX_API_KEY}"
curl -s "https://api.admapix.com/api/data/{endpoint}?{params}" -H "$admapix_auth_header"
curl -s -X POST "https://api.admapix.com/api/data/{endpoint}" \
  -H "$admapix_auth_header" -H "Content-Type: application/json" -d '{...}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The hardcoded external service endpoint means the skill is designed to move data to a third-party API. While this is core functionality, it is still security-relevant because user requests, filters, or business queries may leave the local trust boundary.

Content

Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.

bash
admapix_auth_header="X-API-Key: ${ADMAPIX_API_KEY}"
curl -s "https://api.admapix.com/api/data/{endpoint}?{params}" -H "$admapix_auth_header"
curl -s -X POST "https://api.admapix.com/api/data/{endpoint}" \
  -H "$admapix_auth_header" -H "Content-Type: application/json" -d '{...}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

POST requests to the external API can include structured payloads built from user input, expanding the volume and sensitivity of data transmitted off-platform. In combination with ambiguous triggers, this raises the likelihood of unintentional disclosure to the third party.

Content

Scanner excerpt · SKILL.md (reported line 29)May include surrounding context.

bash
admapix_auth_header="X-API-Key: ${ADMAPIX_API_KEY}"
curl -s "https://api.admapix.com/api/data/{endpoint}?{params}" -H "$admapix_auth_header"
curl -s -X POST "https://api.admapix.com/api/data/{endpoint}" \
  -H "$admapix_auth_header" -H "Content-Type: application/json" -d '{...}'

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Repeating the same ambiguous trigger list reinforces the risk of accidental activation without adding scoping constraints. In a skill that can send user-supplied queries to an external API, imprecise activation increases the chance of unintended data transmission.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill declares only a read tool and describes itself as a thin read-only client, but later claims support for file writing and command execution. This mismatch can mislead a host agent into granting broader execution behavior than intended, increasing the risk of unsafe tool use or privilege expansion through ambiguous instructions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

Detect language from the user's message imposes language selection based on inference rather than explicit user choice. The policy allows locale behavior when the skill offers a language/locale choice or documented justification, which is not present here.

Content

No source excerpt is available for this finding.