skill-improve

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

• Third-party content exposure detected (high risk: 0.75). Required workflow reads session log files from ~/.claude/skill-sessions/<plugin>/sessions/<skill-name>/ at runtime (“Read every file in that directory”), and those logs are outsider-authored free-form text authored by people running prior skill sessions, which is then ingested into the agent’s LLM context for proposal synthesis.