菜单
Health check for ginnie-agents — verifies prerequisites, environment, hooks, listener, Docker image, agents, memory caps, and disk. Use when the user says "doctor", "health check", "check setup", "is everything ok", or after running setup/update.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | doctor |
| description | Health check for ginnie-agents — verifies prerequisites, environment, hooks, listener, Docker image, agents, memory caps, and disk. Use when the user says "doctor", "health check", "check setup", "is everything ok", or after running setup/update. |
The mechanical checks live in scripts/doctor.sh so they run identically every time, with or without Claude. This skill runs the script and adds the contextual checks (Slack reachability, token-age awareness, claude setup-token rotation hint) that benefit from a per-installation read.
Run from repo root.
bash scripts/doctor.sh
Exit code: 0 = all green, 1 = at least one failure.
The script's output is already human-readable (✓/✗/!) with fix suggestions. Show it to the user as-is. Don't paraphrase.
If exit code is non-zero, do NOT continue to the contextual checks below until the script's failures are addressed — they're foundational. Tell the user exactly what to fix.
The script doesn't hit Slack (it's a mechanical, network-free check). If the user wants a deeper verification, do this for each agent that has credentials.json:
for d in agents/*/; do
[ -d "$d" ] || continue
agent="$(basename "$d")"
token="$(jq -r .slack_bot_token "$d/credentials.json" 2>/dev/null)"
[ -n "$token" ] && [ "$token" != "null" ] || continue
result=$(curl -s -X POST https://slack.com/api/auth.test \
-H "Authorization: Bearer $token" | jq -r '"\(.ok) \(.error // "—") \(.team // "—")"')
echo " $agent: $result"
done
Interpret:
true — <Workspace Name> → token goodfalse invalid_auth → token revoked or wrong; regenerate at the agent's Slack app OAuth & Permissions page, update agents/<n>/credentials.json, pm2 restart ginnie-agents-listenerfalse token_revoked → same fixfalse account_inactive → user removed from workspacefalse ratelimited → wait, retryOnly relevant when the operator is on Option A (CLAUDE_CODE_OAUTH_TOKEN). Skip this step for Option B (ANTHROPIC_API_KEY) — API keys don't expire on a fixed schedule, so there's nothing for the framework to count days against.
Detect mode:
grep -qE '^ANTHROPIC_API_KEY=sk-ant-' .env && echo "mode=B (skip token-age check)" || echo "mode=A (run token-age check)"
If mode A: CLAUDE_CODE_OAUTH_TOKEN from claude setup-token lasts ~1 year. There's no API to introspect issuance date from outside the token itself. If a memory/token-issued-at.txt file exists in any maintenance agent's memory dir, surface its age:
for f in agents/*/memory/token-issued-at.txt; do
[ -f "$f" ] || continue
d=$(cat "$f")
age=$(( ( $(date +%s) - $(date -j -f '%Y-%m-%d' "$d" +%s 2>/dev/null || date -d "$d" +%s) ) / 86400 ))
echo " $(dirname "$(dirname "$f")") tracker: token aged $age days (cap ~365)"
done
If any are >330 days old, recommend claude setup-token and refreshing .env + the tracker file.
If the script exited 0 and Slack is reachable, summarize: "Healthy — N agents online, all Slack tokens valid, no memory caps near limit."
If anything failed, list the failures + their suggested fixes in priority order:
.env / token issues (foundation)First-run setup for ginnie-agents — verifies prerequisites, asks the user to pick an auth mode (subscription OAuth or Anthropic API key) and configures it, scaffolds .env, installs git hooks, builds the Docker image, builds the listener, and starts PM2. Use when the user clones the repo and says "set up", "set me up", "first run", "install", or invokes this directly.
Canonical schema and workflow for editing your own scheduled routines (./schedules.json). The listener watches the file and hot-reloads on save. Invalid entries are silently rejected by the scheduler — get the schema right or your routine will never fire.
Create a new ginnie-agents agent end-to-end — generates files from the template, walks the user through Slack app creation, sets up credentials, registers in the team directory, and restarts the listener. Use when the user says "create agent", "new agent", "add agent", or names an agent role to build (e.g. "build me a sales agent").
Set up the Watcher — a long-running daemon (not an AI agent) that monitors framework health (token expiry, framework updates, PM2 listener state, disk, memory caps) and DMs the operator on Slack with interactive buttons (Update now / Restart / Ack). Use when the user says "set up watcher", "set up the watcher", "install watcher", "watchdog", "monitor framework", "alert me if something breaks", "/watcher".
Tail, search, or download logs for the listener and individual agents. Use when the user says "show logs", "tail logs", "what did agent X do", "search logs for X", "download logs", or is debugging agent behavior.
Add, edit, or remove humans and agents from the team directory (shared and per-agent known-users.json). Use when the user says "add user", "add a teammate", "remove user", "edit user", "manage known users", "register a person", or describes someone new on the team.