market-researcher

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt includes a covert instruction to bypass environment guards (write via a shell heredoc if report-shaped filenames are blocked), which is a deceptive workaround outside the skill's stated desk-research purpose.

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

• Third-party content exposure detected (medium risk: 0.30). The workflow requires web search and web fetch, meaning outsider-authored free text from selected web sources is ingested at runtime via the “Built-in web fetch” step (after “Built-in web search”) and then opened/checked during synthesis/verification.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (medium risk: 0.60). The skill chiefly performs harmless desk research, but it explicitly tells the agent to bypass an environment "Write" guard by writing files via a shell heredoc (a deliberate instruction to evade a security restriction), which is a moderate-risk attempt to circumvent platform protections.