@hypercli.com/cli 0.0.0-stage → 0.9.2

package/skills/hypercli-knowledge.json

@@ -0,0 +1,13 @@
+{
+  "name": "hypercli-knowledge",
+  "description": "Upload and inspect platform files, manage durable HyperCLI shared knowledge workspaces, search/sync/download Markdown projections, control grants, and convert captions or documents into local OpenClaw memory. Use for hyper files, hyper workspaces, and hyper memory commands.",
+  "markdown": "---\nname: hypercli-knowledge\ndescription: >\n  Upload and inspect platform files, manage durable HyperCLI shared knowledge\n  workspaces, search/sync/download Markdown projections, control grants, and\n  convert captions or documents into local OpenClaw memory. Use for hyper\n  files, hyper workspaces, and hyper memory commands.\n---\n\n# HyperCLI Knowledge\n\nLoad the `hypercli` and `hypercli-auth` skills before remote operations. Load\nthe `hypercli-flows` skill when a file is only an input to media rendering.\nManaged image references are:\n\n- `/opt/hypercli/docs/cli/commands/files.mdx`\n- `/opt/hypercli/docs/cli/commands/workspaces.mdx`\n- `/opt/hypercli/docs/cli/commands/memory.mdx`\n\n## Choose the storage surface\n\n| Need | Surface |\n| --- | --- |\n| Reusable upload ID for flows/renders | `hyper files` |\n| Durable originals, Markdown projections, search, grants, agent sync | `hyper workspaces` |\n| Local OpenClaw-indexable Markdown generated from source files | `hyper memory import` |\n\nThese are not interchangeable. A platform file ID is not a Workspace path;\nlocal memory import does not upload anything unless `--enrich` calls the model\nAPI, and Workspace `enrich` only builds a payload.\n\n## Platform files\n\nComplete commands: `upload`, `get`, and `delete`.\n\n```bash\nhyper files upload ./reference.png\nhyper files get <file-id>\nhyper files delete <file-id> --yes\n```\n\n`upload` requires a readable local file and returns only once processing makes\nthe upload usable in renders. All three accept the CLI-wide `--json` machine\noutput flag; under `--json`, `delete` skips the confirmation prompt like\n`--yes`. Capture the returned ID carefully. `get` shows state/error/URL. A\nreturned URL may be scoped or time-limited; do not publish it.\n\n`delete` asks for confirmation unless `--yes` (or `--json`); deletion can break\npending or reproducible renders. Resolve the exact ID and check its use before\napproval.\n\n## Workspace command map\n\nAll rows run under `hyper workspaces <command>` (for example\n`hyper workspaces list`).\n\n| Area | Commands |\n| --- | --- |\n| Workspace lifecycle | `create`, `list`, `update`, `delete` |\n| Discovery | `search`, `search-files`, `manifest` |\n| Source lifecycle | `upload`, `wait-until-processed`, `regenerate`, `delete-file` |\n| Retrieval | `download`, `download-url`, `sync` |\n| Access | `grant`, `grants`, `revoke-grant` |\n| Local payload prep | `enrich` |\n\nMost table-oriented commands accept `--output table|json`. `manifest` always\nprints JSON. `sync` uses `--json` rather than `--output json`, and `download`\nuses `--json` for machine-readable local-file metadata.\n\n### Workspace lifecycle and search\n\n```bash\nhyper workspaces create \"Team Knowledge\" --slug team-knowledge --output json\nhyper workspaces list\nhyper workspaces update team-knowledge --description \"Current launch context\"\nhyper workspaces search \"launch notes\" --output json\nhyper workspaces search-files team-knowledge \"visual language\" --no-vector\n```\n\n`update` requires at least one of `--name`, `--slug`, or `--description`.\n`search` spans accessible workspaces; `search-files` is scoped to one. Vector\nsearch is on by default and can be disabled for exact/metadata-only behavior.\nScores and match reasons are relevance signals, not proof that a file contains\nthe answer; retrieve the selected projection before citing it.\n\n`delete <workspace>` is a soft delete but has no confirmation prompt. Confirm\nthe slug/ID and downstream grants/sync consumers first.\n\nExplicit `--user-id` and `--agent-id` select the authorization subject for\nsupported calls and are primarily for agent or local/dev contexts. Never use a\ndifferent subject to bypass access control. When sync receives neither, it\nresolves the authenticated runtime agent or current user.\n\n### Upload and processing\n\n```bash\nhyper workspaces upload team-knowledge ./report.pdf \\\n  --path projects/acme/report.pdf --output json\nhyper workspaces wait-until-processed team-knowledge \\\n  projects/acme/report.pdf --timeout 300 --poll-interval 2\nhyper workspaces manifest team-knowledge\n```\n\nAn upload stores the original and queues its Markdown projection. Retain its\nworkspace-relative path/file ID and wait when the next action requires ready\nMarkdown. A successful upload is not successful conversion. `manifest` shows\nactive projection paths and states.\n\n`regenerate` queues a new Markdown projection and must be followed by\n`wait-until-processed` before consumption:\n\n```bash\nhyper workspaces regenerate team-knowledge projects/acme/report.pdf\nhyper workspaces delete-file team-knowledge projects/acme/obsolete.pdf\n```\n\n`delete-file` soft-deletes without confirmation. Obtain approval and use the\nexact relative path or file ID.\n\n### Download and signed URLs\n\nMarkdown is the default; raw returns the original:\n\n```bash\nhyper workspaces download team-knowledge projects/acme/report.pdf \\\n  --output report.md\nhyper workspaces download team-knowledge projects/acme/report.pdf \\\n  --raw --output report.pdf\nhyper workspaces download-url team-knowledge projects/acme/report.pdf \\\n  --output json\n```\n\n`download` also accepts one `slug/path` argument or reads that address from\nstdin. `--raw` and `--md` are mutually exclusive. Without `--output`, raw uses\nthe source basename; Markdown uses `<source-basename>.md` (for example,\n`report.pdf.md`). Parent directories are created and existing files are\noverwritten, so choose an explicit safe destination.\n\nWith `--json --raw` and no destination, the CLI writes a temporary file and\nprints its `local_path`. JSON does not put raw bytes on stdout.\n\n`download-url` returns a time-limited signed URL for the original. The URL is a\nbearer capability until expiry; never paste it into public logs or durable\ndocuments.\n\n### Sync\n\n```bash\nhyper workspaces sync team-knowledge --output-dir ~/workspaces\nhyper workspaces sync --all --ready-only \\\n  --output-dir ~/workspaces --json\n```\n\nPass either one workspace or `--all`, not both. `--ready-only` avoids writing\nunfinished projections. The deterministic layout is\n`<root>/<workspace-slug>/<source-folder>/.tomd/<source-stem>.md`. Sync writes\nremote Markdown into the local tree; inspect destination ownership and avoid\noverwriting a user's edited copy. Agents commonly sync to\n`/home/node/workspaces` during boot.\n\n### Grants\n\n```bash\nhyper workspaces grant team-knowledge --agent-id <agent> --role viewer\nhyper workspaces grant team-knowledge --user-subject-id <user> --role contributor\nhyper workspaces grants team-knowledge --output json\nhyper workspaces revoke-grant team-knowledge <grant-id>\n```\n\nGrant requires exactly one agent or user subject. Roles are `viewer`,\n`contributor`, and `admin`. Use least privilege, obtain approval, and list\ngrants afterward. Revoke has no confirmation prompt and may break active agent\nsync.\n\n### Enrichment payload\n\n```bash\nhyper workspaces enrich team-knowledge/projects/acme/report.pdf \\\n  --dir ./generated-markdown --json > enrichment.json\n```\n\nThe directory must contain at least one Markdown file. The command recursively\nreads all `.md` files and builds an address/workspace/path/files JSON payload,\nmarking the first sorted file primary. It does not submit that payload to the\nWorkspace API. Review Markdown for secrets before passing the result anywhere.\n\n## Local memory import\n\n`hyper memory` has one command, `import`:\n\n```bash\nhyper memory import ./notes \\\n  --collection research/project \\\n  --workspace ~/.openclaw/workspace\n```\n\nSupported captions are `.srt`, `.vtt`, and `.ttml`; documents are `.pdf`,\n`.doc`, `.docx`, `.epub`, `.txt`, and `.md`. A directory is searched\nrecursively. Install `hypercli-cli[documents]` for PDF/DOCX/EPUB extraction.\nLegacy `.doc` also requires the external `antiword` executable on `PATH`;\nconvert it to DOCX or PDF when that tool is unavailable. Plain text, Markdown,\nand caption conversion do not need optional dependencies.\n\n`--collection` is required and selects a path below the memory root. The root\ndefaults to `<workspace>/memory`, where workspace defaults to\n`~/.openclaw/workspace`; `--memory-dir` or `HYPER_MEMORY_DIR` overrides it.\nThe command prints each generated Markdown path and transcript-line or\nword/character counts.\n\nFor one input, metadata options include `--source-id`, `--title`,\n`--source-url`, caption channel/language/kind/participants, or document author\nand source type. `--source-id`, `--title`, `--source-url`, and `--raw-json3`\nare rejected for multi-file imports; JSON3 applies only to captions.\n`--no-copy-raw` prevents copying original/raw companion files into the memory\nartifact tree.\n\nOptional work changes the trust boundary:\n\n```bash\nhyper memory import ./notes.txt --collection research \\\n  --enrich --model <model-id>\nhyper memory import ./notes --collection research --index --agent <agent-id>\n```\n\n`--enrich` sends sampled extracted text to the remote `/v1` LLM API to produce\nsummary/keyword metadata. Without it, conversion is local. `--index` invokes\nthe local subprocess `openclaw memory index` after every import and fails the\ncommand if indexing fails. Generated files can still exist after that failure;\ndo not rerun import blindly without inspecting them.\n\n## Completion and safety\n\n- Distinguish uploaded, processing, ready, failed, synced, and locally indexed\n  states. They are separate milestones.\n- Treat originals, generated Markdown, search results, grants, signed URLs, and\n  file IDs as potentially private.\n- On a conversion failure, report the exact source and any generated paths;\n  directory imports can partially succeed before a later file fails.\n- Report workspace/file address, processing state, local destination, and\n  whether raw or Markdown content was used. Never expose a signed URL or key.",
+  "commands": [
+    "hyper files",
+    "hyper workspaces",
+    "hyper memory import",
+    "hyper workspaces <command>",
+    "hyper workspaces list",
+    "hyper memory"
+  ]
+}

package/skills/hypercli-voice.json

@@ -0,0 +1,10 @@
+{
+  "name": "hypercli-voice",
+  "description": "Generate speech with HyperCLI reference voice cloning, save audio with --out, and handle voice API errors and identity checks. Use for CLI speech generation, clone reference handling, and voice-capability validation.",
+  "markdown": "---\nname: hypercli-voice\ndescription: >\n  Generate speech with HyperCLI reference voice cloning,\n  save audio with --out, and handle voice API errors and identity checks.\n  Use for CLI speech generation, clone reference handling, and\n  voice-capability validation.\n---\n\n# HyperCLI Voice\n\nLoad the `hypercli-auth` skill before making a remote request. It defines\nthe canonical authentication and credential-safety rules. Read\n`/opt/hypercli/docs/cli/commands/voice.mdx` when exact flags matter, and\nrun `hyper voice clone --help` against the installed\nCLI before relying on remembered options.\n\nSynthesis requires real reference audio. Preset TTS and VoiceDesign commands\nhave been removed; do not fabricate a reference or substitute a named preset.\n\n## Voice Clone\n\n```bash\nhyper voice clone \"Hello world\" --file reference.wav --out cloned.mp3\nhyper voice clone \"Hello world\" --url https://example.com/reference.wav --out cloned.mp3\n```\n\nProvide exactly one reference source: `--file` for local audio or `--url` for\nclient-side HTTPS download before the SDK sends reference bytes. Use `--out` in\nautomated work and whenever `--json` is requested. Treat clone reference audio as\nremote disclosure.\n\n## Preflight\n\n1. Confirm the text, destination path, and that the user expects a remote\n   voice call.\n2. Validate the identity and voice capability with `hyper me`; a valid\n   account without the voice capability will fail.\n3. Treat synthesis text and clone reference audio as remote disclosure,\n   matching the rules in `hypercli-auth`.\n\n## Failure handling\n\n- Missing credentials: run `hyper configure` or set `HYPER_API_KEY`; do not\n  search for, print, or substitute another key.\n- `401` or `403`: stop and report the status and server detail. Retries do\n  not repair authorization — follow `hypercli-auth`.\n- Failed remote call: report whether the destination is absent, complete,\n  or partial before retrying; the server may have completed work after the\n  client lost the response, so do not blindly resubmit.\n- File write failure: choose a writable destination with enough space and\n  an existing or creatable parent directory.\n- Report the output path and whether the result is complete. Never claim\n  success just because the request started.",
+  "commands": [
+    "hyper voice clone --help",
+    "hyper me",
+    "hyper configure"
+  ]
+}

package/skills/hypercli.json

@@ -0,0 +1,23 @@
+{
+  "name": "hypercli",
+  "description": "Operate HyperCLI product APIs, GPU jobs, media flows, voice, and managed agents with the hyper CLI. Use as the router for identity, credentials, agents, compute, flows, files, and voice operations.",
+  "markdown": "---\nname: hypercli\ndescription: >\n  Operate HyperCLI product APIs, GPU jobs, media flows, voice, and managed\n  agents with the hyper CLI. Use as the router for identity, credentials,\n  agents, compute, flows, files, and voice operations.\n---\n\n# HyperCLI\n\nUse the bundled `hyper` command. Run `hyper --help` and\n`hyper <group> --help` before guessing flags. Detailed references live under\n`/opt/hypercli/docs/cli/` in managed images.\n\nFor credential resolution or coding-harness login, load the `hypercli-auth`\nskill. For identity and entitlements load `hypercli-account`. For media\ngeneration load `hypercli-flows`. For GPU jobs load `hypercli-compute`. For\nmanaged runtimes load `hypercli-agents`. For text-to-speech load\n`hypercli-voice`.\n\n## Root surface\n\nThe v1 command tree is deliberately small:\n\n- `hyper me`: identity, capabilities, and entitlements for the active key.\n- `hyper configure`: write the local product API key to\n  `~/.hypercli/config`.\n- `hyper skills`: inventory and print the skills bundled with this CLI.\n- Groups: `hyper agents`, `hyper routines`, `hyper jobs`, `hyper flow`,\n  `hyper files`, `hyper voice`, `hyper websearch`, `hyper llm chat`,\n  `hyper llm image`.\n\nEvery command accepts `--json`; mutations accept `--dry-run` for preflight\nvalidation. Select product/inference with `HYPER_API_BASE` and direct Agent\ncontrol independently with `HYPER_AGENTS_API_BASE` in environment/config.\nWithout an Agents base, Agent endpoints derive from the product base.\nPublic CLIs do not support `--dev` or `--prod`; those\nflags belong to admin tooling. Change the base only when the user intends it.\n\n## Authentication\n\nThe shared SDK/CLI client first uses an explicit canonical credential where\nsupported, then environment `HYPER_API_KEY`, then\nthe canonical key saved by `hyper configure` in the active config file\n(`$HYPER_HOME/config`, or `~/.hypercli/config`). Environment\n`HYPER_AGENTS_API_KEY` is only the final fallback, after all canonical\nsources. Load `hypercli-auth` before changing any of these sources.\n\nNever print, paste, or send a credential. Before a costly or mutating\noperation, validate the intended identity:\n\n```bash\nhyper me --json\n```\n\nIf this returns `401` or says the key is inactive, stop. Do not retry or\nselect another saved key. Report the API base, credential source name, and\nserver detail, then follow the `hypercli-auth` skill. See\n[configuration.mdx](/opt/hypercli/docs/cli/configuration.mdx#diagnosing-401-errors).\n\n## Remote Agents\n\nStart with table output:\n\n```bash\nhyper agents ls\nhyper agents status <agent>\nhyper agents logs <agent>\n```\n\nAdd `-f` to follow logs, and use `hyper agents wait <agent> --state <state>`\nwhen the next step depends on a target state. For lifecycle, routines,\ncontainer access, and runtime configuration, load the `hypercli-agents`\nskill.\n\nDo not paste raw `ls --json` output: it can contain launch environment\nsecrets. Treat log, exec, shell, file-copy, token, config, and routine\noutput as sensitive. Get approval before `exec`, `shell`, `cp` to a remote\ndestination, `agents config` changes, lifecycle mutations, or archive and\nrestore, and dry-run mutations first. Read\n[agents.mdx](/opt/hypercli/docs/cli/commands/agents.mdx).\n\n## Reference Map\n\n- `configure`, `me`, `skills`: `/opt/hypercli/docs/cli/configuration.mdx`\n- `agents`: `/opt/hypercli/docs/cli/commands/agents.mdx`\n- `jobs`: `/opt/hypercli/docs/cli/commands/jobs.mdx`\n- `flow`: `/opt/hypercli/docs/cli/commands/flow.mdx`\n- `files`: `/opt/hypercli/docs/cli/commands/files.mdx`\n- `voice`: `/opt/hypercli/docs/cli/commands/voice.mdx`\n\nIf `/opt/hypercli` is unavailable, rely on command help rather than\ninventing a contract.",
+  "commands": [
+    "hyper",
+    "hyper --help",
+    "hyper <group> --help",
+    "hyper me",
+    "hyper configure",
+    "hyper skills",
+    "hyper agents",
+    "hyper routines",
+    "hyper jobs",
+    "hyper flow",
+    "hyper files",
+    "hyper voice",
+    "hyper websearch",
+    "hyper llm chat",
+    "hyper llm image",
+    "hyper agents wait <agent> --state <state>"
+  ]
+}

package/skills/index.json

@@ -0,0 +1,124 @@
+{
+  "skills": [
+    {
+      "name": "hypercli",
+      "description": "Operate HyperCLI product APIs, GPU jobs, media flows, voice, and managed agents with the hyper CLI. Use as the router for identity, credentials, agents, compute, flows, files, and voice operations.",
+      "commands": [
+        "hyper",
+        "hyper --help",
+        "hyper <group> --help",
+        "hyper me",
+        "hyper configure",
+        "hyper skills",
+        "hyper agents",
+        "hyper routines",
+        "hyper jobs",
+        "hyper flow",
+        "hyper files",
+        "hyper voice",
+        "hyper websearch",
+        "hyper llm chat",
+        "hyper llm image",
+        "hyper agents wait <agent> --state <state>"
+      ]
+    },
+    {
+      "name": "hypercli-account",
+      "description": "Inspect the active HyperCLI identity, capabilities, and entitlements with hyper me, and set up the local product API key with hyper configure. Use for who-am-I, capability, entitlement, and credential-setup questions.",
+      "commands": [
+        "hyper me",
+        "hyper user --json",
+        "hyper configure",
+        "hyper agent onboard",
+        "hyper agent subscribe",
+        "hyper billing balance",
+        "hyper keys list",
+        "hyper keys create",
+        "hyper wallet balance",
+        "hyper wallet topup"
+      ]
+    },
+    {
+      "name": "hypercli-agents",
+      "description": "Create, run, inspect, and control managed HyperCLI agents across the openclaw, hermes, goose, opencode, and buzz runtimes. Use for hyper agents lifecycle and wait, logs, exec, shell, file copies, routines, grant activation with hyper agents activate, dynamic routes, archive and restore, and OpenClaw-only configuration through the hidden hyper agents config and hyper agents models commands.",
+      "commands": [
+        "hyper agents --help",
+        "hyper agents <command> --help",
+        "hyper",
+        "hyper agents wait <agent>",
+        "hyper agents token <agent>",
+        "hyper agents routines create --help",
+        "hyper agents routines update",
+        "hyper me",
+        "hyper agents create <name> --runtime openclaw",
+        "hyper agents config set <agent>",
+        "hyper agents routes list <agent>",
+        "hyper agents routes add <agent> <name> --port N",
+        "hyper agents routes remove <agent> <name>",
+        "hyper agents archive <agent>",
+        "hyper agents restore <agent>",
+        "hyper agents ls"
+      ]
+    },
+    {
+      "name": "hypercli-auth",
+      "description": "Diagnose and configure HyperCLI credentials, API bases, and inactive keys, and authenticate hosted Buzz coding-runtime harnesses. Use when choosing between HYPER_API_KEY and HYPER_AGENTS_API_KEY, running hyper me safely, applying hyper configure, or signing OpenCode, Goose, Claude Code, Codex, or Kimi Code into a hosted runtime.",
+      "commands": [
+        "hyper configure",
+        "hyper me",
+        "hyper me --json",
+        "hyper agents ls --json"
+      ]
+    },
+    {
+      "name": "hypercli-compute",
+      "description": "Browse the live HyperCLI GPU catalog and create, inspect, and control GPU jobs. Use for hyper jobs gpus inventory, job submission preflighted with --dry-run, job list/get/logs, cancellation, and the hidden extend and exec power commands.",
+      "commands": [
+        "hyper jobs --help",
+        "hyper jobs gpus",
+        "hyper jobs create",
+        "hyper jobs list",
+        "hyper jobs get",
+        "hyper jobs logs",
+        "hyper jobs cancel",
+        "hyper jobs extend",
+        "hyper jobs exec",
+        "hyper instances gpus",
+        "hyper instances launch",
+        "hyper launch <image>",
+        "hyper jobs get <job>"
+      ]
+    },
+    {
+      "name": "hypercli-flows",
+      "description": "Create, monitor, wait on, and cancel HyperCLI image and video flows, and upload reusable pipeline inputs with hyper files upload. Use for hyper flow create across the current flow types, flow list/status/wait tracking, and cancelled or failed render handling.",
+      "commands": [
+        "hyper flow --help",
+        "hyper flow create <type> --help",
+        "hyper voice clone",
+        "hyper flow list --json"
+      ]
+    },
+    {
+      "name": "hypercli-knowledge",
+      "description": "Upload and inspect platform files, manage durable HyperCLI shared knowledge workspaces, search/sync/download Markdown projections, control grants, and convert captions or documents into local OpenClaw memory. Use for hyper files, hyper workspaces, and hyper memory commands.",
+      "commands": [
+        "hyper files",
+        "hyper workspaces",
+        "hyper memory import",
+        "hyper workspaces <command>",
+        "hyper workspaces list",
+        "hyper memory"
+      ]
+    },
+    {
+      "name": "hypercli-voice",
+      "description": "Generate speech with HyperCLI reference voice cloning, save audio with --out, and handle voice API errors and identity checks. Use for CLI speech generation, clone reference handling, and voice-capability validation.",
+      "commands": [
+        "hyper voice clone --help",
+        "hyper me",
+        "hyper configure"
+      ]
+    }
+  ]
+}