securitysecrets

mgreenly 的头像作者 mgreenly

秘密管理安全技能

📖 中文导读:秘密管理安全技能 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: security/secrets description: Secrets Management security skill

Secrets Management

API keys and credentials require careful handling throughout their lifecycle.

ikigai Application

API keys (OpenAI, Anthropic, etc.):

  • Store in config file with 0600 permissions
  • Load once at startup, hold in memory
  • Never log, never include in error messages
  • Never embed in source code or commits

Memory handling:

  • Scrub secrets from memory when done: explicit_bzero(key, len)
  • Avoid strdup() for secrets (can't track copies)
  • Keep secret lifetime short and scoped

Config file security:

// Check permissions before reading
struct stat st;
if (stat(path, &st) == 0 && (st.st_mode & 077) != 0) {
    return ERR(ctx, SECURITY, "Config file permissions too open");
}

Never expose:

  • In logs or debug output
  • In error messages shown to user
  • In core dumps (prctl(PR_SET_DUMPABLE, 0))
  • Via environment to child processes

Review red flags: Secrets in printf/logging, strdup on credentials, missing permission checks.