| name | tashan-orgspace |
| description | Install and use the Tashan OrgSpace torg CLI for account, device, organization, capability, and audit operations. Use when a user asks Codex or another agent to work with Tashan OrgSpace, organization accounts, or the torg command. |
Tashan OrgSpace
Use the bundled installer and the torg CLI. Do not reimplement API calls.
Bootstrap the CLI
Resolve paths relative to this SKILL.md file.
- On Windows, stop and explain that this release supports macOS arm64/x64 and Linux x64 only.
- Run
sh scripts/install-cli.sh --check from this Skill directory.
- If the check says this installer has not installed
torg, tell the user that the Skill will install a pinned, checksum-verified CLI in their user directory, then run sh scripts/install-cli.sh --install. The installer tries the official orgspace.tashan.chat source first and starts a fresh GitHub source transaction only after a transport failure; it never falls back after an integrity failure.
- Never run the installer with
sudo, copy it into a shell startup file, or set its test-only environment variables.
- If installation succeeds but
torg is not on PATH, invoke $HOME/.local/bin/torg and tell the user to add $HOME/.local/bin to PATH later.
An installed or updated Skill is discovered by Codex on the next turn. Do not claim it was active earlier in the same turn.
Establish the server contract
For agent-driven calls, include --invocation-source ai_via_cli and prefer --json:
torg --invocation-source ai_via_cli --json health
torg --invocation-source ai_via_cli --json capability list
torg --invocation-source ai_via_cli --json capability describe <capability-id>
Inspect live capabilities before using a feature; the bundled capability list is a coverage gate, not proof that the production server is currently reachable. If health or discovery fails, report that fact and stop rather than inventing support.
Authenticate safely
Read authentication.md before registering, logging in, resetting a password, or changing devices. Let torg collect passwords and verification codes through its hidden interactive prompt. Never place a password, code, access token, refresh token, or credential-file passphrase in arguments, environment variables, chat, logs, or command output.
Use the login token stored by the CLI as the real user and device identity. Do not create subaccounts to represent additional computers.
Execute commands
- Start with
<command> --help when flags or targets are uncertain.
- Read with
--json and use the returned IDs exactly; do not guess organization, account, device, cursor, or capability IDs.
- Preserve CLI confirmation gates. Add
--yes only after the user has explicitly requested the mutation and you have summarized its exact target.
- Supply a fresh opaque
--idempotency-key for mutating commands that accept it. Reuse that key only to retry the same intended mutation.
- Do not bypass authorization or reach into credential storage. See safety.md.
Examples:
torg --invocation-source ai_via_cli --json auth whoami
torg --invocation-source ai_via_cli --json org list
torg --invocation-source ai_via_cli --json org member list --org <org-id>
torg --invocation-source ai_via_cli --json device list
torg --invocation-source ai_via_cli --json audit list --limit 25
The prerelease currently covers authentication (including verification-code SMS), device sessions, organizations, capability discovery, and audit reads. Files, OKR/tasks, approvals, chat, remote execution, services, databases, domains, general notification SMS, and AI employees remain planned capabilities unless the live capability endpoint says otherwise.
Respond
Report the command outcome, affected resource IDs, and any server error code without exposing secrets. Distinguish CLI installation success from production service readiness.
