菜单
Index vanilla DayZ source (Enforce Script, .layout, config.cpp, .rvmat) into a per-user vector database so DayZ agents can do semantic search via the dayz-rag MCP server. One-time setup; rerun with --full after a DayZ update. Required before agents can use the search_dayz_source tool.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Audit and fix DayZ .p3d model files for collision, action targeting, physics, animation, path, and structural issues. Runs py3d-based validation against known-working reference patterns. Also validates config.cpp and .rvmat files for path and property issues. Use this skill whenever: a DayZ mod object has no collision, actions don't appear, the player walks through placed objects, CCTObject raycasts miss, Geometry LOD seems ignored, a model was exported from Blender and doesn't work in-game, textures are missing, animations don't play, or the user says "review the p3d", "check the model", "audit collision", "why no actions", "flag/object has no collision", "audit the mod", "check paths", "validate the model". Also trigger when debugging any placed Inventory_Base item. This is the GO-TO skill for P3D and mod structure debugging.
Bootstrap Community-Online-Tools (COT) admin permissions on a DayZ test server in one shot. Auto-prepends @CF + @Community-Online-Tools to the mod chain, launches the server + client, waits for your character to spawn (detected when COT writes the per-player files), kills the session, grants you the `admin` role in your Players JSON, flips every ` 0` to ` 2` in your personal `Permissions/<id>.txt` file, then relaunches. Per-player grant only — `Roles/everyone.txt` is never touched. Requires /dayz-init + /dayz-add-server + /dayz-build-pbo first.
Pull the latest Agentic-Z template improvements (agents, skills, conventions, docs) from upstream into your clone, without touching your DayZ mod work in workspace/. Adds the upstream remote on first run, fetches, shows a changelog, then merges only template-managed paths. Re-runs sync-skills automatically. Refuses if your working tree is dirty unless --force is passed.
Wipe ALL template-managed artifacts across every domain so the repo is ready to push. Orchestrator — discovers and runs every domain's `<domain>-clean-workspace` skill in turn (currently just `dayz-clean-workspace`). Future domains plug in here automatically when they ship their own cleanup skill following the naming convention. Interactive confirmation by default; --yes skips it; --dry-run lists what each domain would clean.
Add missing DayZ mod scaffolding pieces (config.cpp stub, $PBOPREFIX$, workbench/dayz.gproj + DayZSetting.xml, README, .gitkeeps) into an EXISTING workspace/<ModName>/ folder. Idempotent - never overwrites files that are already present. Used by /dayz-import-mod after symlinking an external mod, but also runnable directly when you cloned a mod into workspace/ manually and want the standard skeleton filled in.
Set up a DayZ test server instance under .server/<instance>/. Copies the mission template from DayZ Server install if missing, creates per-instance serverDZ.cfg + Profiles/ (server -profiles=) and a top-level shared .server/!ClientDiagLogs/ (client -profiles=). Each instance is isolated on the server side so you can run multiple variants of the same map (chernarus, chernarus-hardcore, etc.) without cross-contamination. Refuses if the legacy workspace/_server/ layout exists (delete it manually; that layout is no longer supported). Required before /dayz-launch-test for a given instance. Use --refresh-mission to re-copy mission content after a DayZ update.
| name | dayz-search-index |
| description | Index vanilla DayZ source (Enforce Script, .layout, config.cpp, .rvmat) into a per-user vector database so DayZ agents can do semantic search via the dayz-rag MCP server. One-time setup; rerun with --full after a DayZ update. Required before agents can use the search_dayz_source tool. |
Path note:
<skill-dir>in commands below is the absolute path of this skill's folder. When the agent loads this skill the harness exposes the skill's base directory; substitute it before running. Sibling skills are reached via<skill-dir>\..\dayz-X\.
Build a semantic-search index over vanilla DayZ source so DayZ specialist agents stop flying blind through P:\ with Grep. Backs the dayz-rag MCP server.
Follow .claude/skills/_shared/dayz-conventions.md.
| Source | Files | Chunking |
|---|---|---|
P:\scripts\**\*.c | Enforce Script (~2,800 files) | per class/method |
P:\gui\**\*.layout | UI layouts (~210 files) | per top-level Widget block |
P:\dz\**\config.cpp | Vanilla configs (~200 files) | per class X { ... } block |
P:\**\*.rvmat | Materials (~35,000 files) | whole file, content-deduped by SHA-256 |
P:\**\*.xml | XML (types, etc.) | batched 50 elements/chunk |
model.cfg is intentionally skipped — vanilla extracted data doesn't ship them.
~/.claude/dayz-search-index/ (= C:\Users\<you>\.claude\dayz-search-index\)
lancedb/ — vector storemanifest.json — embed model used, per-source chunk counts, indexed-at, token usage + $ estimateconfig.json — model name the MCP server should use at query timePer-user, gitignored (lives outside the repo). Survives clones.
Embedding runs through Voyage AI's hosted API. Code-tuned model, much higher retrieval quality than local CPU options for our DayZ corpus. The trade-off is a network call per batch and an API key.
Default model: voyage-code-3 — code-tuned, 32k context, 1024-dim. $0.18 per 1M tokens with 200M tokens free (Series-4 free allowance, doesn't expire when payment method is added).
Cost reality: a full vanilla rebuild is ~65M tokens (measured, not estimated — dominated by ~34k unique rvmats and ~10k XML chunks from types.xml batches). At voyage-code-3's $0.18/M that's ~$12, but the 200M free tier covers ~3 rebuilds. After that, switch to voyage-4-lite ($0.02/M ≈ $1.30/rebuild) via VOYAGE_MODEL env var if cost matters. Live token + $ counter prints during the embed loop, and final totals are saved to manifest.json.
Free-tier projection check: before kicking off the embed phase, the skill estimates this build's tokens (sum of chunk text lengths ÷ chars-per-token) and projects what the cumulative monthly usage would be after this run. If the projection exceeds 80% of the 200M free tier, you get a prompt to confirm. Cumulative usage is tracked locally in ~/.claude/dayz-search-index/usage.log (one JSONL entry per build). Caveats: the local counter only sees builds run via these skills — it does NOT see usage by other apps sharing your VOYAGE_API_KEY. The authoritative usage view is at dash.voyageai.com. Free tier resets monthly per Voyage's billing cycle. Pass --ignore-tier-warning to skip the prompt (CI / scripted use). Override the budget with VOYAGE_FREE_TIER_TOKENS=<int> env var if you're on a non-default plan.
.env at the repo root (gitignored — never commits):
VOYAGE_API_KEY=pa-xxxxxxxxxxxx
VOYAGE_MODEL=voyage-4-lite # cheapest ($0.02/M, 16M TPM)
VOYAGE_MODEL=voyage-code-3 # default, code-tuned
VOYAGE_MODEL=voyage-4-large # best general quality
The indexer auto-installs voyageai + python-dotenv on first run.
First time / after DayZ update:
python "<skill-dir>\index.py" --full
Status only:
python "<skill-dir>\index.py" --status
P:\ may have changed).DayZ RAG indexer
[OK] P:\ mounted
[OK] Vanilla data: P:\dz
[OK] VOYAGE_API_KEY loaded (pa-GRJ...edF)
[INFO] Index: C:\Users\you\.claude\dayz-search-index\
[INFO] Embedding model: voyage-code-3 (1024D, via Voyage cloud)
[INFO] Pricing: $0.18 per 1M tokens (Series-4 includes 200M free)
Walking vanilla folders: scripts, gui, dz, core, graphics, languagecore
scripts/ c=28432c/2805f
gui/ layout=1842c/214f
dz/ cpp=3201c/206f
core/ ...
rvmat dedup: 33800 duplicates (97.6%)
Embedding 34322 chunks via Voyage (voyage-code-3, document mode)...
embedding: 100%|██| 34322/34322 [03:42<00:00] 12,847,231 tok / ~$2.312
[OK] Index built: 34322 chunks
[INFO] Tokens used: 12,847,231 | Estimated cost: $2.3125
[INFO] Embed time: 222.4s
VOYAGE_API_KEY is missing — no silent partial runs.--force (cost/runtime backstop).pip install -r requirements.txt (the indexer auto-installs on first run if missing).
lancedb — file-based vector storetqdm — progress barsvoyageai — Voyage embedding SDKpython-dotenv — load VOYAGE_API_KEY from .env/dayz-preflight's warnings (Tools, Workshop) — only P:\ mounted is mandatory.~/.claude/dayz-search-index/..env. It's in .gitignore; keep it there.--full — the LanceDB table is fixed-dim and switching providers/models will mismatch query and document vectors.