flags-sdk
Audited by ZeroLeaks on Apr 15, 2026
This skill is flagged AT_RISK due to two concrete concerns. First, transparency is weakened by what appears to be a hardcoded credential and an obfuscated or encoded execution path in SKILL.md, making it harder to fully review what the skill actually does before loading it. Second, and more critically, the skill weakens instruction/data boundaries by treating untrusted external content as policy-level input, which materially increases prompt-injection risk compared to a no-skill baseline. Behavior analysis was not run, so downstream behavioral impact remains unvalidated, but the prompt-injection finding alone is sufficient to warrant caution—this skill does materially worsen the agent's security posture in tested scenarios.
The skill has 2 transparency concerns that weaken pre-use reviewability, mainly around potential hardcoded credential and obfuscated or encoded execution path.
The skill increases prompt injection risk by weakening trust boundaries around untrusted external content treated as policy.
Behavior analysis was not run.
Potential hardcoded credential
Untrusted external content treated as policy
Obfuscated or encoded execution path