continuous-learning-v2

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

• Third-party content exposure detected (high risk: 0.80). The skill's CLI (scripts/instinct-cli.py, cmd_import) explicitly fetches arbitrary http/https URLs via urllib.request.urlopen and parses/imports that untrusted web content into local "instinct" files that can influence agent behavior (import → instincts → evolve/promote), so external webpages could inject instructions indirectly.