happier-diagnose

# Happier Diagnose Investigate a Happier issue from real evidence (logs, doctor output, source), determine the root cause at high confidence, and present findings. Then — only with explicit user consent — upload a private diagnostics bundle to Happier developers **and/or** open a sanitized public GitHub issue. The two paths are complementary: the private bundle gives maintainers raw artifacts; the public issue gives them (and the community) a searchable, durable record. Doing both is normal. Mostly hands-off; minimize questions to the user. ## Evidence rules (read first) - **High confidence only.** Never guess. Every claim in the final root cause must be traceable to a specific log line, a `happier doctor` field, or a specific source file (with line numbers). If you cannot reach high confidence, say "Inconclusive — here's what I observed and what's missing" rather than inventing a story. - **Symptom ≠ root cause.** A failed RPC call is a symptom; the daemon being stale, the access key being missing, or a provider returning 429 is a root cause. - **No fabricated paths or flags.** If a path, file, or command is not in this skill, in the user's logs, or in code you have actually read, do not include it. - **Concrete code references are welcomed.** The Happier maintainers prefer issues that name specific files (`apps/ui/sources/.../AgentInput.tsx:1759`), include code snippets, and propose a hypothesis. This is repo-relative — different from the user's filesystem paths, which m...