메뉴
Provision, use, collect, and tear down multi-endpoint libcrafter lab sessions across Hetzner, QEMU, VirtualBox, and future lab providers. Use when oracle, probe, or generated tools need provider-backed packet work that must be independent of the developer machine.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Provision and use one disposable remote endpoint provider target for libcrafter packet work. Use for single-endpoint raw sends, captures, debugging, artifact collection, or endpoint teardown; use lab-session for coordinated multi-endpoint oracle/probe workflows.
Add or maintain libcrafter lab provider adapters, capability contracts, provider matrix tests, and documentation for substrate-independent oracle and probe execution.
Add or run libcrafter packet behavior validation through oracle specs, backend adapters, offline checks, pcap checks, live checks, and artifacts. Use when packet behavior changes need validation coverage.
Open, update, title, review, or prepare libcrafter pull requests so commits are ready for rebase-and-fast-forward landing and required validation is documented.
Use ignored .scratch workspaces for experiments, disposable prototypes, network probes, QEMU or dongle labs, generated exploratory code, and any task where the user asks to make a scratch folder or wants code/scripts that should not be committed until promoted.
Create, amend, reword, split, squash, or prepare libcrafter commits with the repository's Conventional Commits policy and scoped staging discipline.
| name | lab-session |
| description | Provision, use, collect, and tear down multi-endpoint libcrafter lab sessions across Hetzner, QEMU, VirtualBox, and future lab providers. Use when oracle, probe, or generated tools need provider-backed packet work that must be independent of the developer machine. |
Use this skill when a task needs a disposable provider-backed session with two or more coordinated endpoints. A lab session owns substrate setup, repository push/bootstrap, endpoint metadata, artifact collection, and teardown; oracle, probe, or generated tools own the workload they run inside that substrate.
For one-off manual work on a single endpoint, use endpoint-provider. For adding
or changing provider adapters, use lab-provider.
tools/lab/run plan --provider <provider> --dry-run --profile smoke --seed 1 --role stimulus --role targettools/oracle/run live --provider <provider> --dry-run --profile smoke --seed 1 --count 10tools/probe/run --provider <provider> --dry-run --profile smoke --seed 1 --count 10--confirm-live-run.Supported lab providers are registered in tools/lab/engine/providers/.
Use provider names from the lab registry instead of hard-coded provider lists.
Provider-backed oracle and probe runs should go through lab sessions so packet
exchange behavior is independent of the substrate where the agent is running.
Lab-backed appliance execution uses coarse runtime profiles. Use lan-raw for
private VM or lab roles, wan-raw for WAN-capable roles, whad-serial for
prepared WHAD serial dongle roles, and dot11-monitor for prepared
monitor-mode Wi-Fi roles. Do not request per-protocol endpoint capabilities
when an appliance profile describes the placement requirement.
For reusable hardware-backed roles, acquire a persistent endpoint asset lease by appliance profile before the run, and release it after artifacts are collected. Real hardware identifiers, SSIDs, MACs, public addresses, credentials, and sensitive captures stay in ignored local state.
Do not store credentials, provider account identifiers, public IPs, live host IDs, or packet captures from sensitive networks in tracked files. If credentials or local virtualization prerequisites are missing, keep dry-run artifacts and report the skipped live work clearly.
Keep the artifacts needed to debug a failed run offline:
Tracked documentation may describe artifact shapes, but live artifacts belong
under ignored output directories such as target/ or lab state/artifact roots.