meta-ads

Warn

Audited by Snyk on Mar 12, 2026

Risk Level: MEDIUM
Full Analysis

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and ingests public third‑party content — e.g., campaign/scripts/lp_analyzer.py scrapes arbitrary landing page URLs, campaign/scripts/ad_library.py pulls/scrapes the Meta Ad Library (and generates web research prompts), and campaign/references/stage-prompts.md (Stage 4) tells the sub-agent to use web_search/web_fetch on Reddit, forums and review sites — and those findings are written to market-research.md / customer-profile.md and fed into downstream stages (strategy, creative generation, uploads), so untrusted external content can materially influence agent decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (high risk: 0.90). The skill launches Claude/Opus sub-agents (campaign/references/stage-prompts.md and ad_library.py) that instruct the model to run web_search/web_fetch against external sites (e.g., reddit / site:facebook.com/ads/library / site:adspy.com) at runtime and inject those fetched pages/quotes into the agent's context (market-research.md / competitor-ads.md), so external URLs like site:facebook.com/ads/library and reddit.com are used at runtime to directly control prompts and outputs.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

  • Direct money access detected (high risk: 1.00). The skill explicitly integrates with the Meta (Facebook) Graph API requiring permissions such as ads_management and includes scripts and functionality that modify live campaigns: auto_optimize.py ("auto-pause bleeders, shift budget to winners"), ad_uploader.py / upload_to_meta.py ("Upload to Meta Ads Manager"), and a Meta Graph API wrapper (meta_api.py). Those components indicate the skill programmatically updates ad campaigns and budgets (i.e., manages ad spend) via the Meta Ads API, which is a specific financial-execution capability per the policy (Managing Ad Spend Budgets via API). Other integrations (Genviral image generation) are non-financial and irrelevant to this decision.

Issues (3)

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W012
MEDIUM

Unverifiable external dependency detected (runtime URL that controls agent).

W009
MEDIUM

Direct money access capability detected (payment gateways, crypto, banking).

Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 12, 2026, 07:28 PM
Issues
3