grok-host-prep
Skill2 starsChanged 40 days agoMITView on GitHub
One-time (re-runnable) host prep for Grok Build: check binaries, lsp.json, [features] lsp_tools, MCP command existence, and efficiency knobs. Use when setting up a machine, enabling LSP, or a Feed harness_gap is mcp / permissions / 8192. Triggers: grok-host-prep, /grok-host-prep, prepare machine, setup LSP, tweak grok host, host prep.
- Reads credentials
In this repository49 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adversarial-auditSkill
- agentic-reactorSkill
- agent-orchestratorSkill
- ai-optimizationSkill
- architecture-synthesisSkill
- audit-allow-buildsSkill
- audit-ide-dependenciesSkill
- author-workflow-skillSkill
- bdd-strategizerSkill
- chrome-extension-mv3Skill
- concurrent-cli-agentsSkill
- control-feederSkill
- control-graphSkill
- cursor-transcript-harvestSkill
- cv-promote-guardSkill
- devcontainer-hardenedSkill
- dual-copy-skill-publishSkill
- explore-repo-readonlySkill
- finder-reactorSkill
- fix-dependency-securitySkill
- fusion-sageSkill
- git-worktreesSkill
- gt-flowSkill
- higher-order-decision-architectSkill
- looperSkill
- master-plannerSkill
- mvu-refactor-planSkill
- odysseus-navigatorSkill
- project-editor-profileSkill
- react-client-expertSkill
- semantic-markup-cssSkill
- semantic-nameSkill
- session-unit-orderSkill
- shell-kernel-ontologySkill
- skill-rename-propagationSkill
- split-to-prsSkill
- src-tree-reorganizeSkill
- stellar-roadmapSkill
- stellar-spacemapSkill
- subagent-delegationSkill
- subagent-explore-reportSkill
- supply-chain-hardenSkill
- tauri-agenticSkill
- tauri-ipc-debugSkill
- tidy-commit-pushSkill
- upgrade-packagesSkill
- verification-cockpitSkill
- x-agent-resourcesSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
Run the script (check first):
- Distinctive: line 37
If the report is what you want on disk:
- Very specific: line 73
Do not treat permission_mode = always-approve as a prep win (EVA forbids it).
- Very specific: line 79
No secrets printed; no grok nested
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
grok-host-prep
Load rule: This skill owns host checks + safe apply. Prompt routing stays in control-feeder. Do not paste MCP env/secrets. Do not nest
grok.
Check : default — report OK / WARN / MISS
Apply : write missing lsp.json servers; append [features] lsp_tools if absent
Install: optional LSPs only with --installWhen to use
| Use | Skip |
New host, missing lsp tool, Feed harness_gap | Mid-turn code edit |
| Want rust-analyzer / clangd / tsserver wired | Changing models or permission_mode |
MCP command might be missing from PATH | Dumping or editing MCP secrets |
Steps
- Run the script (check first):
bash ~/.grok/skills/grok-host-prep/scripts/prepare.sh --check- If the report is what you want on disk:
bash ~/.grok/skills/grok-host-prep/scripts/prepare.sh --apply- Optional LSPs (only if you want them globally):
bash ~/.grok/skills/grok-host-prep/scripts/prepare.sh --apply --install- New Grok session (or restart) so
lsp.json+lsp_toolsload. - Session flags the script will not write into your shell rc — set if you want the extra tools:
| Flag | Effect |
GROK_LSP_TOOLS=1 | same as [features] lsp_tools = true |
GROK_WEB_FETCH=1 | enable web_fetch (off by default) |
GROK_SUBAGENTS=1 | enable task / subagents |
GROK_MEMORY=1 | expose memory_search / memory_get (config [memory] enabled is separate) |
What actually makes this host faster
Owned by the harness, not a second graph:
| Knob | Why |
~/.grok/lsp.json + lsp_tools | goToDefinition / references / hover instead of fat read_file |
codebase_indexing (default true) | code-graph context; leave on |
output_byte_limit ≥ 8192 | do not lower (repo AGENTS.md) |
rg + rust-analyzer + clangd on PATH | this tree is Rust + some C |
MCP command exists | dead [mcp_servers.*] wastes startup; project .grok/config.toml replaces a name entirely |
Official grok = ~/.grok/bin/grok | leftover ~/.local/bin/grok can steal PATH |
Do not treat permission_mode = always-approve as a prep win (EVA forbids it).
Done when
- Script printed a full check table
--apply(if requested) wrote only missing LSP servers andlsp_tools = true- No secrets printed; no
groknested
Do not
- Rewrite
config.tomlwholesale or print[mcp_servers.*.env] - Lower
output_byte_limit --installwithout the user asking for extra LSPs- Invent a second agent OS / extra graph crate
Related
Skills: control-feeder (Feed harness_gap) · eva-emptiness (always-approve is not a prep win)
Plugin: eva-emptiness (auth tether). Official grok binary + bundled/ after login — not this script.
Notes in grok-build intelli-arch-designs/:
| Note | Use from this skill |
harness-feed-filter.md | harness knobs (perms, 8192, AGENTS.md, MCP) ≠ graph |
harness-economy-model.md | tools dominate $; do not lower bash below 8192 or polish skill-list |
harness-learning-cost-roadmap.md | official grok; measure with tool-mix-observe; no second OS |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

