Agent Skills
› opensquilla/opensquilla
› meta-long-running-build-watchdog
meta-long-running-build-watchdog
GitHub已弃用的元技能，用于通过tmux监控长时间运行的构建任务，并允许子代理自动诊断失败并应用修复。因存在安全风险且缺乏回滚机制，目前处于禁用状态，不建议重新启用。
Trigger Scenarios
用户请求监控长时间运行的构建任务（如模型微调、CI构建）
需要自动诊断间歇性失败的回归测试套件
Install
npx skills add opensquilla/opensquilla --skill meta-long-running-build-watchdog -g -y
SKILL.md
Frontmatter
{
"kind": "meta",
"name": "meta-long-running-build-watchdog",
"always": false,
"metadata": {
"opensquilla": {
"risk": "high",
"capabilities": [
"shell",
"tmux",
"filesystem-write",
"subprocess"
]
}
},
"triggers": [],
"provenance": {
"origin": "opensquilla-original",
"license": "Apache-2.0"
},
"composition": {
"steps": [
{
"id": "launch",
"with": {
"task": "Start a detached tmux session running the command described in: {{ inputs.user_message | xml_escape | truncate(512) }}"
},
"skill": "tmux"
},
{
"id": "inspect",
"with": {
"task": "After a short interval, scrape the pane output of the session started above and report any error or warning lines."
},
"skill": "tmux",
"depends_on": [
"launch"
]
},
{
"id": "heal",
"with": {
"task": "Diagnose the captured logs and propose \/ apply a fix. Logs: {{ outputs.inspect }}"
},
"skill": "sub-agent",
"depends_on": [
"inspect"
]
},
{
"id": "memorize",
"kind": "tool_call",
"tool": "memory_save",
"tool_args": {
"mode": "append",
"path": "memory\/build-watchdog.md",
"content": "{{ outputs.heal }}"
},
"depends_on": [
"heal"
],
"tool_allowlist": [
"memory_save"
]
}
]
},
"description": "[DEPRECATED] Build watchdog — launches arbitrary commands from the user message in tmux and lets sub-agent auto-apply a fix. Disabled pending the E5 bounded sub-agent contract + Jinja sandbox + side-effect ledger (plan §3.1 A1\/A8 \/ §5.3 E4): the launch task interpolates raw user_message into a shell-bound tmux session and the heal step lets sub-agent mutate state with no rollback. Do not re-enable without `metadata.opensquilla.risk: high` + capabilities {shell, tmux, filesystem-write, subprocess} and a saga-style compensation step.",
"meta_priority": 0,
"disable-model-invocation": true
}
Long-Running Build Watchdog (Meta-Skill)
Watches a long-running command via tmux, lets sub-agent diagnose
failures and propose a fix, and records the diagnosis to memory.
Designed for overnight model fine-tunes, CI image builds, or repeated
regression suites that may fail intermittently.
Fallback
Manually start a tmux session, scrape output, ask the LLM to diagnose, record the resolution.
Version History
- 7f72a32 Current 2026-07-05 18:41
