process-logs

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires the agent to read raw error messages, trpc_input/attempted_value and other log fields and embed them verbatim into Beads task titles, Task(...) prompts, and DB notes — which would cause any secrets present in logs (API keys, tokens, passwords) to be output exactly.

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md requires querying /admin/logs via SQL (error_logs.error_message and generation_trace.error_data->>'message') and then delegating to subagents with prompts that include those free-text error messages/stack traces, so outsider-authored text can enter the LLM context through runtime database log contents.