ワンクリックで
code-review
// Use when: reviewing or modifying hiredis-happ C++ code, Redis/hiredis async lifecycle, cluster routing, command ownership, tests, or CI risk.
// Use when: reviewing or modifying hiredis-happ C++ code, Redis/hiredis async lifecycle, cluster routing, command ownership, tests, or CI risk.
Use when: running hiredis-happ tests, adding regression coverage, debugging CTest failures, or planning Redis/raw/cluster integration tests.
Use when: updating AGENTS.md, CLAUDE.md, compatibility shims, Agent Skills, source indexes, or cross-tool AI configuration for hiredis-happ.
Use when: configuring, building, packaging, installing, or debugging CMake/CI/dependency/runtime-path issues in hiredis-happ.
| name | code-review |
| description | Use when: reviewing or modifying hiredis-happ C++ code, Redis/hiredis async lifecycle, cluster routing, command ownership, tests, or CI risk. |
Use this skill for code review, risk triage, and correctness checks in hiredis-happ.
AGENTS.md, the relevant header in include/detail/, and the matching implementation in src/detail/.doc/ai/source-index.md before making semantic changes.doc/ROADMAP.md.cmd_exec must be callback-called and destroyed exactly once.redisReply* beyond the hiredis callback.reply, reply->str, reply->element, and redisAsyncContext* before dereferencing.redisAsyncConnect() returns a non-null context with c->err, free it with redisAsyncFree(c).redisAsyncDisconnect() and pending callbacks as asynchronous lifecycle boundaries; verify with integration tests before refactoring release paths.va_copy() must have a matching va_end().{...} pair only.MOVED: update routing and retry; empty endpoint reuses current IP with the redirected port.ASK: send ASKING first and retry only the current command without permanently updating the slot map.TRYAGAIN: use bounded retry through the existing TTL/timer path.CLUSTER SLOTS: validate reply type, element pointers, slot range, host strings, and incomplete coverage.bin in PATH before tests.git diff --check; the repository has CRLF-aware whitespace rules.Report what was verified, what changed, what tests ran, and what remains risky. Avoid vague claims such as "should be safe" without test or source backing.