How do I install the Infra Bootstrap skill?

Run npx skillmds@latest add gleb-nikitin/infra-bootstrap in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Infra Bootstrap safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: executes scripts. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Infra Bootstrap?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Infra Bootstrap free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Infra Bootstrap?

gleb-nikitin (@gleb-nikitin) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Infra Bootstrap by gleb-nikitin — Agent Skill

Name: Infra Bootstrap
Rating: 4.2 (5 reviews)
Author: gleb-nikitin

Infra Bootstrap

Deterministic onboarding for a new Vast.ai GPU server used in SDXL training. Use when Codex must run explicit preflight/bootstrap with a single JSON config: connect to server, verify GPU, ensure remote directories, upload one dataset, ensure base model file, validate Python imports, and write a concise machine-readable report.

gleb-nikitin Updated Sep 22, 2026

File contents

Infra Bootstrap (v2)

Default flow is v2 and requires explicit config. Old heuristic flow is legacy and is not default.

Trigger Conditions

Use this skill when task is onboarding/preflight of a new Vast.ai GPU server for SDXL.

Input Contract

Pass one JSON file with explicit fields:

server: host, port, ssh_key
remote_dirs: list of directories to mkdir -p
dataset: local_path, remote_path
model: local_path, remote_path
python_checks: python_bin, required_imports

Example schema: references/config.example.json

Commands

Validate config only:

cd /Users/glebnikitin/Projects/CodexSkills
./skills/infra-bootstrap/scripts/run validate-config --config ./skills/infra-bootstrap/references/config.example.json

Run end-to-end v2:

cd /Users/glebnikitin/Projects/CodexSkills
./skills/infra-bootstrap/scripts/run setup --config ./skills/infra-bootstrap/references/config.example.json

Legacy flow (not default):

cd /Users/glebnikitin/Projects/CodexSkills
./skills/infra-bootstrap/scripts/run legacy setup --dataset <name> --server <server_name>

Safety and Idempotency

Safe to run repeatedly.
Default mode never uses delete sync behavior.
--destructive is required to enable rsync --delete for dataset sync.
Fail-fast: report includes failed_step and error message.

Artifact Source Policy

Production bootstrap uses ready artifacts from Storage Box / persistent storage only.
This skill does not include model download from Google Drive as part of standard flow.
Any Google Drive downloader scripts are legacy/pre-prod fallback tools and must be used outside normal bootstrap runs.

Report Format

setup writes one JSON report (default: ./logs/infra/bootstrap_v2_<timestamp>_<runid>.json):

version, status, started_utc, finished_utc
config, destructive
steps[]: {step, status, details}
on failure: failed_step, error

Troubleshooting

validate_config: fix missing paths/fields in config.
connect_server: verify SSH host/port/key and network reachability.
verify_gpu: ensure NVIDIA runtime is enabled and nvidia-smi works on remote.
validate_python_imports: install missing Python packages in remote environment.

---name: infra-bootstrapdescription: Deterministic onboarding for a new Vast.ai GPU server used in SDXL training. Use when Codex must run explicit preflight/bootstrap with a single JSON config: connect to server, verify GPU, ensure remote directories, upload one dataset, ensure base model file, validate Python imports, and write a concise machine-readable report.---# Infra Bootstrap (v2)Default flow is `v2` and requires explicit config.Old heuristic flow is `legacy` and is not default.## Trigger ConditionsUse this skill when task is onboarding/preflight of a new Vast.ai GPU server for SDXL.## Input ContractPass one JSON file with explicit fields:- `server`: `host`, `port`, `ssh_key`- `remote_dirs`: list of directories to `mkdir -p`- `dataset`: `local_path`, `remote_path`- `model`: `local_path`, `remote_path`- `python_checks`: `python_bin`, `required_imports`Example schema: `references/config.example.json`## CommandsValidate config only:```bashcd /Users/glebnikitin/Projects/CodexSkills./skills/infra-bootstrap/scripts/run validate-config --config ./skills/infra-bootstrap/references/config.example.json```Run end-to-end v2:```bashcd /Users/glebnikitin/Projects/CodexSkills./skills/infra-bootstrap/scripts/run setup --config ./skills/infra-bootstrap/references/config.example.json```Legacy flow (not default):```bashcd /Users/glebnikitin/Projects/CodexSkills./skills/infra-bootstrap/scripts/run legacy setup --dataset <name> --server <server_name>```## Safety and Idempotency- Safe to run repeatedly.- Default mode never uses delete sync behavior.- `--destructive` is required to enable `rsync --delete` for dataset sync.- Fail-fast: report includes `failed_step` and error message.## Artifact Source Policy- Production bootstrap uses ready artifacts from Storage Box / persistent storage only.- This skill does not include model download from Google Drive as part of standard flow.- Any Google Drive downloader scripts are legacy/pre-prod fallback tools and must be used outside normal bootstrap runs.## Report Format`setup` writes one JSON report (default: `./logs/infra/bootstrap_v2_<timestamp>_<runid>.json`):- `version`, `status`, `started_utc`, `finished_utc`- `config`, `destructive`- `steps[]`: `{step, status, details}`- on failure: `failed_step`, `error`## Troubleshooting- `validate_config`: fix missing paths/fields in config.- `connect_server`: verify SSH host/port/key and network reachability.- `verify_gpu`: ensure NVIDIA runtime is enabled and `nvidia-smi` works on remote.- `validate_python_imports`: install missing Python packages in remote environment.

Frequently asked questions