メニュー
Spawn AI subagents (claude, pi, copilot) in a new tmux window. Trigger on requests containing 'subagent', 'spawn', 'spin up', or 'spin out'. Use weight keywords (heavy/medium/light) for preset agent/model/effort/yolo combos, or specify explicit agent + model + effort + yolo options.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when interacting with Kubernetes clusters, pods, deployments, services, namespaces, helm releases, or any k8s resource. Triggers on kubectl, helm, k8s, pod, deployment, namespace, etc.
Use when interacting with Linear.app from the command line. Triggers on any request involving Linear issues, projects, cycles, sprints, teams, roadmaps, milestones, documents, triage, notifications, comments, bulk operations, or any Linear workflow. Also triggers when the user references a Linear issue identifier (e.g. ENG-123, LIN-456), wants to list/create/update/close/assign issues, check sprint status, run burndown charts, import/export Linear data, manage webhooks, or automate Linear tasks from the terminal. Use proactively whenever the user mentions "Linear", "linear issue", "linear sprint", "linear project", "linear ticket", or any Linear-related workflow — even if they don't explicitly ask for the CLI.
Use when addressing, responding to, or resolving pull request review comments and inline code suggestions. Triggers whenever the user wants to work through PR feedback — even if they say 'address the review comments', 'fix the PR feedback', 'apply the suggestions', 'resolve the threads', 'respond to reviewer comments', or just 'let's work through this PR review'. Use proactively after a PR review is received.
Security testing skill for Nuclei — a fast, template-driven vulnerability scanner by ProjectDiscovery. Use this skill to: scan URLs, APIs, or OpenAPI specs for vulnerabilities; run or write nuclei YAML templates and workflows; generate templates from a STRIDE threat model report; or run a full pipeline (threat model → templates → scan). Triggers on: "run nuclei", "scan this URL", "nuclei template", "nuclei workflow", "nuclei from threat model", "web app security scan", "pen test this API", "full security pipeline", "threat model and scan", SQLi, XSS, SSRF, IDOR, auth bypass, CORS misconfig. Use proactively when the user shares a URL, OpenAPI spec, source code, or STRIDE report and wants security testing — even if they haven't mentioned nuclei by name.
Queries Panther SIEM data lake via GraphQL API. Use when the user wants to query Panther logs, run SQL against the data lake, search for indicators of compromise, explore databases/tables/columns, poll for query results, list past queries, build GraphQL scripts, or manage Panther alerts. Triggers on: "query Panther", "Panther data lake", "Panther GraphQL", "search Panther logs", "indicator search in Panther", "executeDataLakeQuery", "executeIndicatorSearchQuery", "panther_logs", "Panther alerts", "list alerts", "alert status", "triage alert", "resolve alert", "assign alert", "mark alert as resolved", "update alert", "comment on alert", or any request to pull log data, hunt threats, or manage security alerts via the Panther API. Also triggers on "search my logs", "find this IP in Panther", "show me my alerts", or "close these alerts".
Interact with the Snyk REST API (v2025-11-05) for vulnerability management, security scanning, and developer security. Use for: listing orgs/projects/targets, querying issues (package vulnerabilities, licenses, cloud misconfigs, SAST), searching audit logs, managing service accounts, memberships, invitations, fetching SBOMs, and looking up package vulnerabilities by PURL. Requires SNYK_TOKEN injected via 1Password `op run`.
| name | subagent |
| description | Spawn AI subagents (claude, pi, copilot) in a new tmux window. Trigger on requests containing 'subagent', 'spawn', 'spin up', or 'spin out'. Use weight keywords (heavy/medium/light) for preset agent/model/effort/yolo combos, or specify explicit agent + model + effort + yolo options. |
| metadata | {"author":"Peter Benjamin","version":"0.1.0"} |
| user-invocable | true |
Spawn an AI agent as a detached interactive session in a new tmux window.
When the user asks for a subagent, parse the request into:
claude, pi, or copilot (copilot maps to pi).--model.low, medium, high, xhigh, max for --effort.yolo or
dangerously-skip-permissions, add --dangerously-skip-permissions.~/Projects/github.com/komodohealth/, or the current
working directory if another is not implied.to, for, or the main verb.Then run:
./scripts/spawn.py \
--agent <claude|pi|copilot> \
--model <model> \
[--effort <level>] \
[--yolo] \
--cwd <path> \
--prompt "<task>"
Use these defaults when the user gives a weight keyword without explicit agent/model/effort:
| Weight | Agent | Model | Effort | Yolo |
|---|---|---|---|---|
heavy | claude | opus | medium | yes |
medium | pi | github-copilot/kimi-k2.7 | — | no |
light | pi | github-copilot/gpt-5.4-mini | — | no |
User: spin up a heavy subagent to do xyz
./scripts/spawn.py --agent claude --model opus --effort medium --yolo --cwd ~/Projects/github.com/komodohealth/ --prompt "do xyz"tmux window running
claude --model opus --effort medium --dangerously-skip-permissions "do xyz"User: spawn a light subagent to do foobar
./scripts/spawn.py --agent pi --model github-copilot/gpt-5.4-mini --cwd ~/Projects/github.com/komodohealth/ --prompt "do foobar"tmux window running
pi --model github-copilot/gpt-5.4-mini "do foobar"spin up a claude subagent with sonnet 5 and high reasoning in yolo mode to develop a new feature
./scripts/spawn.py --agent claude --model sonnet-5 --effort high --yolo --cwd ~/Projects/github.com/komodohealth/ --prompt "develop a new feature"tmux window running
claude --model sonnet-5 --effort high --dangerously-skip-permissions "develop a new feature"./scripts/spawn.py relative to this skill directory.