@anythingai/cli 0.0.1

package/package.json

@@ -0,0 +1,59 @@
+{
+  "name": "@anythingai/cli",
+  "version": "0.0.1",
+  "homepage": "https://anything.com/",
+  "license": "MIT",
+  "bin": {
+    "anything": "./dist/js/bin.mjs"
+  },
+  "files": [
+    "dist",
+    "skills"
+  ],
+  "exports": {
+    "./bin": "./dist/js/bin.mjs"
+  },
+  "publishConfig": {
+    "access": "public",
+    "registry": "https://registry.npmjs.org"
+  },
+  "scripts": {
+    "build": "tsdown --no-dts --out-dir dist/js",
+    "prepublishOnly": "yarn build",
+    "test:unit": "jest --config=jest.unit.config.ts --passWithNoTests",
+    "test:ci": "jest --config=jest.unit.config.ts --coverage --reporters=default --reporters=github-actions --maxWorkers=2 --passWithNoTests",
+    "test:integration": "jest --config=jest.integration.config.ts --passWithNoTests",
+    "typecheck": "tsgo --noEmit"
+  },
+  "dependencies": {
+    "conf": "^15.1.0",
+    "graphql": "^16.12.0",
+    "graphql-ws": "^6.0.4",
+    "neverthrow": "^8.2.0",
+    "ws": "^8.18.0",
+    "yargs": "^17.7.2",
+    "zod": "^3.24.1"
+  },
+  "devDependencies": {
+    "@createinc/env": "^0.0.0",
+    "@createinc/flux-api": "0.0.0",
+    "@createinc/tsconfig": "0.0.0",
+    "@jest/environment": "^30.2.0",
+    "@swc/core": "^1.9.3",
+    "@swc/jest": "^0.2.37",
+    "@types/jest": "^30.0.0",
+    "@types/ws": "^8.18.0",
+    "@types/yargs": "^17.0.33",
+    "@typescript/native-preview": "^7.0.0-dev.20260113.1",
+    "dotenv-flow": "^4.1.0",
+    "jest": "^30.2.0",
+    "jest-environment-node": "^30.2.0",
+    "reflect-metadata": "^0.2.2",
+    "testcontainers": "^10.18.0",
+    "tsdown": "^0.20.3",
+    "typescript": "^5.9.3"
+  },
+  "engines": {
+    "node": ">=24.0.0"
+  }
+}

package/skills/anything-cli/SKILL.md

@@ -0,0 +1,233 @@
+---
+name: anything-cli
+description: Use when an agent needs to build, inspect, or manage an Anything app from the CLI. Covers auth, account and org checks, project creation, follow-up generations, secrets, publishing, domains, databases, deployments, members, and compound workflows. Best for OpenClaw, Claude Code, Codex, Cursor, or any agent that should drive Anything non-interactively.
+allowed-tools:
+  - Bash(anything:*)
+  - Read
+---
+
+# Anything CLI
+
+Use `anything` as the control-plane CLI for Anything apps. Prefer this skill when the user wants to build or manage an Anything app without depending on the web UI.
+
+## Core Rules
+
+- Prefer `ANYTHING_API_KEY` for non-interactive auth. Use `anything auth login --api-key <key>` for key-based login.
+- Start with `anything user --json` for account, org, plan, and credit checks.
+- Use the resource-first shape everywhere: `anything projects ...`, `anything domains ...`.
+- **Always use `--json`** when another tool or agent will read the output. The JSON envelope format is: `{ "ok": true, "command": "...", "data": { ... }, "meta": { "duration_ms": 142 } }`.
+- **Use `--quiet`** to get only the primary ID or URL for chaining: `PROJECT_ID="$(anything projects create --quiet --prompt "...")"`
+- Capture `projectId` from create output and pass it explicitly on later commands.
+- Pass flags for all inputs. Do not rely on prompts or hidden local state.
+- `--non-interactive` is auto-enabled when `CLAUDECODE`, `CODEX`, `CI`, or `OPENCLAW` env vars are detected.
+- Use `--dry-run` on state-modifying commands to preview actions before executing.
+- If the user asks for something outside the known surface area, run `anything introspect` or `anything <subcommand> --help`.
+
+## Output Modes
+
+| Flag                | Behavior                                                                                                         |
+| ------------------- | ---------------------------------------------------------------------------------------------------------------- |
+| `--json`            | Returns JSON envelope: `{ ok, command, data, meta }` or `{ ok: false, command, error: { code, message, hint } }` |
+| `--quiet`           | Prints only the primary identifier (ID or URL) — one per line for list commands                                  |
+| `--dry-run`         | Outputs planned actions as JSON without executing                                                                |
+| `--non-interactive` | Hard-fails instead of prompting; emits `ActionRequired` JSON to stderr                                           |
+
+## Exit Codes
+
+| Code | Meaning            | Agent action                               |
+| ---- | ------------------ | ------------------------------------------ |
+| 0    | Success            | Continue                                   |
+| 1    | General error      | Inspect error                              |
+| 2    | Invalid arguments  | Fix command syntax                         |
+| 3    | Resource not found | Check ID                                   |
+| 4    | Auth failure       | Re-authenticate                            |
+| 5    | Conflict           | Resource exists                            |
+| 6    | Timeout            | Retry (also used for publish wait timeout) |
+| 7    | Server error       | Retry with backoff                         |
+
+## Quick Start
+
+```bash
+anything user --json
+anything projects create --org <org-id> --prompt "Build a CRM for a roofing company" --json
+anything projects generate <project-id> --prompt "Add auth and billing" --json
+anything projects messages <project-id> --json
+anything projects get <project-id> --json
+anything projects publish <project-id> --json          # waits for deploy
+```
+
+## One-Shot Workflow
+
+```bash
+anything ship --prompt "Build a todo app with auth" --json
+```
+
+The `ship` command creates (or generates on existing), then publishes and waits for the deployment to finish. It streams progress as NDJSON in `--json` mode, or `[init]`/`[progress]`/`[done]` markers in text mode. Use `--no-wait` for fire-and-forget workflows.
+
+```bash
+anything ship --project pg_123 --prompt "Add a settings page" --json
+anything ship --project pg_123 --skip-publish --prompt "Add dark mode"
+anything ship --no-wait --prompt "Build a landing page" --json
+```
+
+## Workflow
+
+1. Run `anything user --json`.
+2. Pick the recommended org if one is available. Otherwise use the org the user specifies.
+3. Create the project and store `projectId`.
+4. Iterate with `projects generate`, `projects messages`, and `projects get`.
+5. Add secrets before publish when the app needs third-party credentials.
+6. Publish once the project is ready.
+
+## Command Reference
+
+### Account and org discovery
+
+```bash
+anything user --json
+anything orgs --json
+anything orgs get <org-id> --json
+anything orgs members <org-id> --json
+anything status --json
+anything switch                                    # interactive org/project picker
+anything llm-context --json                        # combined agent context
+anything introspect                                # full command tree as JSON
+anything skill                                     # print SKILL.md
+```
+
+### Project creation and iteration
+
+```bash
+anything projects create --org <org-id> --prompt "<prompt>" --json
+anything projects generate <project-id> --prompt "<prompt>" --json
+anything projects messages <project-id> --limit 20 --json
+anything projects get <project-id> --json
+anything projects files list <project-id> --json
+anything projects files get <project-id> app/page.tsx --json
+anything projects rename <project-id> --name "New Name" --json
+anything projects duplicate <project-id> --json
+```
+
+### Secrets and publish
+
+```bash
+anything projects secrets add <project-id> --name "OPENAI_API_KEY" --value "$OPENAI_API_KEY" --json
+anything projects secrets list <project-id> --json
+anything projects publish <project-id> --json                       # waits for deploy
+anything projects publish <project-id> --no-wait --json             # fire-and-forget
+anything projects publish status <project-id> <deployment-id> --json
+anything projects unpublish <project-id> --yes --json
+```
+
+### Domains
+
+```bash
+anything domains list <org-id> --json
+anything domains add app.example.com --project pg_123 --json
+anything domains remove dom_123 --yes --json
+anything domains verify dom_123 --json
+```
+
+### Databases
+
+```bash
+anything databases list --org <org-id> --json
+anything databases get <database-id> --json
+anything databases create --project pg_123 --json
+anything databases query <database-id> "SELECT * FROM users LIMIT 5" --json
+anything databases connect <database-id>
+anything databases reset <database-id> --yes --json
+```
+
+### Deployments
+
+```bash
+anything deployments list <project-id> --json
+anything deployments get <deployment-id> --json
+anything deployments logs <deployment-id> --json
+anything deployments rollback <project-id> --yes --json
+anything deployments rollback <project-id> <deployment-id> --yes --json
+```
+
+### Members
+
+```bash
+anything members list --org <org-id> --json
+anything members invite user@example.com --org <org-id> --json
+anything members remove user@example.com --org <org-id> --yes --json
+anything members role user@example.com admin --org <org-id> --json
+```
+
+### Assets
+
+```bash
+anything assets upload <project-id> ./logo.png --json
+anything assets list <project-id> --json
+anything assets remove <project-id> <asset-id> --json
+```
+
+### Logs (with streaming)
+
+```bash
+anything projects logs <project-id> --json
+anything projects logs <project-id> --since 1h --json
+anything projects logs <project-id> --follow --json            # NDJSON stream
+anything projects logs <project-id> --follow --level error     # tail errors only
+```
+
+### Project context
+
+```bash
+anything projects set <project-id>          # set active project
+anything projects unset                     # clear active project
+anything orgs set <org-id>                  # set active org
+anything orgs unset                         # clear active org
+anything link <project-id>                  # link directory to project (.anything/)
+anything unlink                             # remove directory link
+anything pull                               # pull files + secret names
+anything pull --files                       # pull project files only
+anything pull --env                         # write secret names to .env.example
+```
+
+### Watch
+
+```bash
+anything watch <project-id>                 # stream all events
+anything watch <project-id> --events generation
+anything watch <project-id> --json          # stream events as NDJSON
+```
+
+## Chaining with --quiet
+
+```bash
+PROJECT_ID="$(anything projects create --quiet --prompt "Build a todo app")"
+anything projects generate "$PROJECT_ID" --prompt "Add dark mode" --quiet
+PUBLISH_URL="$(anything projects publish "$PROJECT_ID" --quiet)"
+echo "Published to: $PUBLISH_URL"
+```
+
+## Dry Run
+
+```bash
+anything projects create --dry-run --org org_1 --prompt "test" --json
+# Returns: { ok: true, command: "projects create", dry_run: true, planned_actions: [...] }
+
+anything projects generate pg_123 --dry-run --prompt "Add auth" --json
+anything projects rename pg_123 --dry-run --name "New Name" --json
+anything databases create --dry-run --project pg_123 --json
+anything domains add app.example.com --dry-run --project pg_123 --json
+anything deployments rollback pg_123 --dry-run --json
+```
+
+## Failure Handling
+
+- Missing auth: set `ANYTHING_API_KEY` or run `anything auth login --api-key <key>`.
+- Missing org context: rerun `anything user --json` and choose an explicit org ID.
+- Zero credits: switch to an org with credits or ask the user to top up before generating.
+- Command uncertainty: run `anything introspect` or `anything <subcommand> --help`.
+- Exit code 6 (timeout): retry the command.
+- Exit code 7 (server error): retry with exponential backoff.
+
+## References
+
+- For agent-specific installation in OpenClaw, Claude Code, Codex, and Cursor, see `references/setup.md`.

package/skills/anything-cli/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: 'Anything CLI'
+  short_description: 'Agent guide for building apps with Anything CLI'
+  default_prompt: 'Use $anything-cli to build or manage an Anything app from the CLI.'

package/skills/anything-cli/references/setup.md

@@ -0,0 +1,66 @@
+# Setup
+
+This skill ships with `@anythingai/cli`. Keep the skill identifier as `anything-cli`, but all command examples should invoke the public binary name `anything`.
+
+## Find the packaged skill
+
+After installing `@anythingai/cli`, locate the packaged skill directory and copy it into the agent runtime you are configuring.
+
+For a global npm install:
+
+```bash
+GLOBAL_NPM_ROOT="$(npm root -g)"
+SKILL_DIR="$GLOBAL_NPM_ROOT/@anythingai/cli/skills/anything-cli"
+```
+
+For local repo development:
+
+```bash
+SKILL_DIR="/absolute/path/to/escher/apps/anything-cli/skills/anything-cli"
+```
+
+## OpenClaw
+
+```bash
+mkdir -p ~/skills
+cp -R "$SKILL_DIR" ~/skills/anything-cli
+```
+
+## Claude Code
+
+```bash
+mkdir -p ~/.claude/skills
+cp -R "$SKILL_DIR" ~/.claude/skills/anything-cli
+```
+
+## Codex
+
+```bash
+mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
+cp -R "$SKILL_DIR" "${CODEX_HOME:-$HOME/.codex}/skills/anything-cli"
+```
+
+## Cursor
+
+Cursor supports project rules in `.cursor/rules/*.mdc` and a root `AGENTS.md` file. The most portable setup is to paste the packaged skill guidance into one of those two places.
+
+### Option 1: project rule
+
+Create `.cursor/rules/anything-cli.mdc`:
+
+```md
+---
+description: Use Anything CLI to build and manage Anything apps non-interactively
+alwaysApply: false
+---
+
+[Paste the body of SKILL.md here]
+```
+
+### Option 2: root AGENTS.md
+
+Append the skill guidance to the repo's root `AGENTS.md` under a section like `Anything CLI`.
+
+## Auth reminder
+
+For public agent setups, prefer `ANYTHING_API_KEY`. Keep `ANYTHING_API_URL` for local or staging overrides only.