Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/chanmeng666/archlang/archlangnpx skills add ChanMeng666/archlang --skill archlanggit clone --depth 1 https://github.com/ChanMeng666/archlangWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/chanmeng666/archlang/archlang)<a href="https://agentmods.dev/skills/chanmeng666/archlang/archlang"><img src="https://agentmods.dev/badge/skills/chanmeng666/archlang/archlang.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00104 | $0.05717 |
| Opus 5 | $0.00052 | $0.02858 |
| Sonnet 5 | $0.00021 | $0.01143 |
| Haiku 4.5 | $0.00010 | $0.00572 |
Grade A, and why
archlang scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 308 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ArchLang — author floor plans as code
ArchLang turns a small .arch text file into a professional floor-plan drawing. It is built for
agents: deterministic, self-correcting (errors carry a machine code, a prose fix, and often a
machine-applicable fix arch fix can apply), and verifiable without ever looking at an image
(arch describe).
Setup (zero-install)
The CLI runs straight from npm — no clone, no build:
npx @chanmeng666/archlang help
(Or npm i -g @chanmeng666/archlang to get a persistent arch binary.)
The loop (always follow this)
- Learn the language first. Run
arch specand read it — the entire language in one page (~2k tokens). (arch contextprints everything: spec + this workflow + CLI reference + error catalog.) Do this before writing any.arch. - Write the plan to a
.archfile (or pipe via stdin with-), preferring the placement sugar below so you never hand-compute a coordinate. - Render it:
arch compile plan.arch -o plan.svg --json. The JSON is{ ok, diagnostics, summary }. - Auto-fix the mechanical faults: if
okis false, runarch fix plan.arch --dry-run --jsonto preview the machine-applicable edits (off-wall opening → attachment form, out-of-range position clamped, …), then re-run without--dry-runto apply. Anythingfixcan't resolve stays indiagnostics[].fixfor you to edit by hand. Exit code2means a deterministic user error — fix it, don't blindly retry (1= IO/internal,3= bad usage). - See the plan without an image:
arch compile plan.arch -f txt(orarch preview plan.arch --ascii) prints a zero-dependency ASCII floor plan you can read straight from stdout. - Verify intent:
arch describe plan.arch --jsonreturns the rooms (areas, adjacency), what each door connects, and totals. Confirm the room count, labels, and areas match what was asked. - Gate on soundness — don't ship a flagged plan.
arch validate plan.arch --strict --json(parse + resolve + lint).--strictmakes every advisory warning fail (exit2) — the gate a generation pipeline runs before it ships. Add--graph g.jsonto also assert the intended room-to-room adjacency ({ "living": ["kitchen","hall"], … }); a mismatch fails. Read eachdiagnostics[].fix, edit, and re-run until it passes — or, if a warning is deliberate, say so. - Check the plan against the brief. Write the user's brief as an
intent.json— its checkable expectations as data (room count, per-room concepts with area/window bands, total area, optional adjacency/reachability). Two disciplines keep it brief-grounded: assert an area band only where the brief gives a number ("about/~N" → ±10%; "at least N" →minonly; qualitative words → nothing), and assert the top-level roomcountonly when the brief enumerates the rooms. Gate witharch validate plan.arch --intent intent.json --feedback --json: a gating miss (room count/existence/area/total-area/window) fails (exit2) with a per-violation correction prompt — iterate on the feedback and re-run. Adjacency/reachability are advisory (reported, never fail the gate). Usearch score plan.arch --brief intent.json --jsonas a continuous satisfaction meter (always exit0) to watch the plan approach the brief across edits. See/intent.schema.json. - Fix furniture geometry:
arch repair plan.arch -o fixed.archpushes furniture out of walls/doorways/swing arcs (the geometric corrector; distinct fromfix). It rewrites a piece in the form you wrote it (anatpoint, or theinsetof anin <room> anchor …placement) — and always readunresolved: a piece it may not rewrite (onefor/component statement drawing several pieces, expression coordinates, anagainst wallanchor) is reported there with the fault and why, never silently skipped.changed: falsewith a non-emptyunresolvedmeans you edit the source. - Show the user:
arch preview plan.arch -o plan.pngrenders a PNG (--installfetches the optional renderer if missing).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today Changed · +9 lines c1305b450274
- yesterday Changed · +38 lines 6d1d7d6ad667
- 5d ago First seen · 261 lines · 104 tokens per session scan A e15c5d26aed8
archlang is a skill published in the GitHub repository ChanMeng666/archlang (8 stars, last pushed today), licensed MIT. It adds 104 tokens to every session and 5,717 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
review-architecture
Review a PR against the Pascal architectural rules — package boundaries (core/viewer/editor/nodes), the registry-driven composition model (def.geometry / def.renderer / def.system), legacy-dispatch regressions, the slots + world-scale-UV convention for new nodes/geometry, hook hygiene (useEditor/useScene/useViewer)…
open-pr2
Open or update a pull request on pascalorg/editor with a plain-language issue-and-fix description based on the full branch diff. Use only when the user explicitly asks for OpenPR2 or /open-pr2.
open-pr
Open a pull request on pascalorg/editor using the repo's PR template. Use when the user asks to open/create a PR, push and PR, or ship a branch in the editor repo.
codebase-onboarding
Use when you land in an unfamiliar or inherited codebase and must get productive fast: a breadth-first map of entry points, request flow, module ownership, hidden side effects (cron, webhooks, workers) and churn hotspots, committed as CODEBASE-MAP.md. NOT a deep audit of one module (that is analyze) or chasing one…
github-multi-repo
Multi-repository coordination, synchronization, and architecture management with AI swarm orchestration.
subsystem-map
This skill should be used before answering "does Genesis have X", "does Genesis lack X", auditing Genesis capabilities, comparing Genesis to an external system, or reviewing/summarizing the architecture. It routes to the canonical judgment-layer subsystem map so audits start from the map, not from a cold grep. Also…