kai-ad-campaign

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs extracting API credentials from .env.local and then inserting them into curl requests (e.g., access_token=${META_TOKEN}), which requires reading and embedding secret values into generated commands/requests and creates an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). In E.1 Pull Active Campaigns the required workflow calls the Meta Graph API to list active campaigns and ad-level insights (ids/names/creative fields) from the marketer’s ad account, which are first-party provider data but include user-side/outside-generated ad creatives and text that the LLM ingests at runtime.

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

• Direct money access detected (high risk: 1.00). The skill explicitly includes Meta Ads API execution: it shows extracting credentials and curl calls to query accounts/ads, and Phase 5 details an API execution flow that creates campaigns and ad sets (including targeting + budget), creates ads, and discusses bid/budget writes and activation. Those are concrete API actions to manage ad spend (i.e., update budgets/activate campaigns), which qualifies as Direct Financial Execution authority under the policy.