Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/CraftOS-dev/CraftBotnpx agentmods add skills/craftos-dev/craftbot/agent-app-creatorWrote 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/craftos-dev/craftbot/agent-app-creator)<a href="https://agentmods.dev/skills/craftos-dev/craftbot/agent-app-creator"><img src="https://agentmods.dev/badge/skills/craftos-dev/craftbot/agent-app-creator/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/craftos-dev/craftbot/agent-app-creator"><img src="https://agentmods.dev/badge/skills/craftos-dev/craftbot/agent-app-creator.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00038 | $0.04717 |
| Opus 5.5 | $0.00015 | $0.01887 |
| Sonnet 5.5 | $0.00008 | $0.00943 |
| Haiku 4.5 | $0.00004 | $0.00472 |
Grade A, and why
agent-app-creator 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 23d ago.
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 — 361 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent App Creator
A Agent App is a self-contained local web app: one PocketBase process (data, auth, realtime, custom verbs) serving a React frontend built from a preset kit. You declare schema, compose UI, wire verbs — the platform owns the rest.
Step 0: Have a registered project (MANDATORY FIRST)
- Task instruction contains
Project ID+Project Path→ the project is already scaffolded. Use those values. Skip scaffolding. - No Project ID in your instruction (user asked in a regular chat) → call
agent_app_scaffold(name, description, auth_mode)— it scaffolds AND dispatches the build to the project's dedicated session. Tell the user the build started, then end your turn. Do NOT build in the chat session.
Pick auth_mode from requirements: none (personal local tool — default) or
multi-user (accounts; the kit's LoginGate wraps the app automatically).
The ownership rule (the gate enforces this)
Edit ONLY:
| Path | Purpose |
|---|---|
frontend/src/app/ |
all UI code |
pb/pb_migrations/ |
schema — one NEW migration per change |
pb/pb_hooks/ops.pb.js + new *.pb.js / *.js modules |
custom verbs + their helpers |
operations.json |
declarations for those verbs (non-system entries) |
AGENT_APP.md |
your plan/context/index — keep current |
NEVER edit frontend/src/kit/, frontend/src/main.tsx, frontend/src/config.gen.ts,
pb/pb_hooks/_system.pb.js, manifest.json, or build configs — the validation
gate hashes them and fails the build if they changed. Need a variant of a
kit component? Wrap it in app/:
// frontend/src/app/components/DueBadge.tsx
import { cn } from '../../kit/index.ts';
export function DueBadge({ overdue }: { overdue: boolean }) { /* compose */ }
Before coding
- If
reference/requirements.mdstarts withMARKETPLACE DECISION: install <app-id>— do NOT build. Callagent_app_marketplace_install(app_id=..., name=..., description=..., will_adapt=<true if the decision line says adapt: yes>). It installs INTO this project (same tab and id — never a duplicate).adapt: no— the install completes the build and the system announces it; do NOT send your own summary and do NOT call notify_ready or walk_verify. End the run.adapt: yes— after the install, apply ONLY the adaptations listed under## Adaptations(modify flow: edit →agent_app_notify_ready→agent_app_walk_verify). If the list says "none specified", ask the user what to change (a FINALsend_message) instead of guessing. The user explicitly chose reuse over a fresh build — never rebuild what was just installed, even if a later trigger asks you to "continue" it.
- Read
agent_file_system/GLOBAL_AGENT_APP.md— colors, fonts, enforced rules. - Read
{project_path}/AGENT_APP.mdandreference/requirements.md. The creation wizard interviewed the user and synthesizedrequirements.md— it is the binding spec: implement it exactly and mirror its checklist intoAGENT_APP.md. If it is absent, build from the project description; only ask the user (a FINALsend_message,continue_work=false) when something is genuinely blocking and you cannot reasonably decide it yourself. - Any feature need data from outside the app? Check, then research.
FIRST check the
[INTEGRATIONS this app can use]block already in your context — if a connected integration's action covers the feature (email =send_gmail), usebridge.callAction; nothing to research. Only for THIRD-PARTY public APIs: research like an engineer — endpoint, auth, response shape, limits. Spawn a research_agent; never write an integration hook from memory.- User named an API/service → research it. If it needs a key, tenant URL,
or account detail you cannot find online, ask the user (final
send_message) and build the rest of the app while waiting. - No API named → research candidates and pick a keyless public API yourself (e.g. Open-Meteo for weather). Choosing the source is your engineering call — no user round-trip.
- Nothing usable exists → build the honest empty/offline state and REPORT the blocker in your final message. Mock or generated data is forbidden unless requirements explicitly ask for demo data.
- User named an API/service → research it. If it needs a key, tenant URL,
or account detail you cannot find online, ask the user (final
- A Agent App build is substantial work — the standard run protocol applies
as-is (scope, plan, execute, verify, deliver); this skill adds nothing to
it.
reference/requirements.mdis the binding spec verification checks against; mirror the feature checklist inAGENT_APP.md.
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 23d ago First seen · 361 lines · 38 tokens per session scan A 93982e7a2e66
agent-app-creator is a skill published in the GitHub repository CraftOS-dev/CraftBot (395 stars, last pushed today), licensed MIT. It adds 38 tokens to every session and 4,717 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-24.
Other skills, from other repositories
mem0-oss-to-platform
Plan, then execute, a migration of a project from the mem0 open-source / self-hosted SDK (local Memory class) to the mem0 Platform / hosted SDK (MemoryClient). TRIGGER when: the developer wants to move, switch, or migrate mem0 off OSS/self-hosted to the hosted API, e.g. "migrate my mem0 setup to the platform", "switch…
fastapi-router-py
Create FastAPI routers with CRUD operations, authentication dependencies, and proper response models. Use when building REST API endpoints, creating new routes, implementing CRUD operations, or adding authenticated endpoints in FastAPI applications.
dev-runtime
How the Node side of a dev stack runs without haven, and the backend host: pnpm dev (UI built once, backend reloading), pnpm dev:hmr (adds the Vite ui lane), tools/dev-runtime (api + worker in one Node process), backend reload on a file change, the debounce (quiet window + max wait), ports derived from PORT. Use when…
migrate-segw-to-rap
Reverse-engineer a SEGW-built OData V2 service (MPC/DPC/MPCEXT/DPCEXT) into a modern RAP V4 service — tables, CDS views (interface + projection), behavior definitions, draft entities, service definition + binding. Use when asked to "migrate this SEGW service to RAP", "convert OData V2 to V4 RAP", "modernize this…
telnyx-messaging-hosted-curl
Set up hosted SMS numbers, toll-free verification, and RCS messaging. Use when migrating numbers or enabling rich messaging features. This skill provides REST API (curl) examples.
api-design
Use when settling the contract of an API you expose, before implementation: resources/URLs, REST vs GraphQL, versioning, one RFC 9457 error envelope, pagination, idempotency — emitted as OpenAPI 3.1. NOT implementing the endpoints (that is fastapi/nestjs/go/nodejs), NOT auth hardening (that is secure-coding), NOT…