compensation-strategy

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The "Companion check" explicitly instructs the agent to run local shell commands (ls ~/.claude/skills/ ...) and offer/install packages via npx, which is unrelated to compensation advice and constitutes hidden operational instructions that probe/modify the local environment.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md "Benchmark a role against market" core task explicitly instructs the agent to "pull data from at least two sources" and lists public, user-generated sites (e.g., Levels.fyi, Glassdoor, LinkedIn Salary), meaning the agent is expected to fetch and interpret untrusted third-party market data that can directly influence compensation decisions and follow-up actions.