What this skill does
Creates a Salesforce
trial org by inserting a
sObject with the
CLI — the same request the Trialforce/developer
web signup form issues under the hood. There is no bespoke signup endpoint: signup = inserting a
(key prefix
) against an authenticated
host org (a Trialforce Source Org / Env Hub / partner org that is entitled to create trial orgs).
The flow is
two CLI calls: (1) create the
, (2) read it back to pick up the assigned org id (creation is asynchronous — the org id appears shortly after the insert).
Prerequisites — confirm before executing
-
Authenticated host org — always confirm which one, explicitly. You are signing up
from an authenticated host org, not anonymously. The
CLI operates against an org you have already logged into once (
, or
).
Always pass an explicit () with the host org's alias or username on every command, and confirm the target with the user before creating — even if a default org is configured. Creating a
is a real provisioning action; do not let it run against whatever org happens to be the default.
- Do not rely on the default-org fallback. With no , the CLI resolves the target from → env var → local then global config, and errors () if none is set — it never auto-picks among your connected orgs. That default may be an unrelated dev/scratch org, so an omitted is either wrong-org or a hard failure. Never omit it.
- The user may have many authenticated orgs. Run and, if the intended host org is ambiguous or not provided, ask the user which alias/username to use. Do not guess.
- Verify the chosen org is in before creating (stale refresh tokens / expired certs show as error states, not ).
- Never invent credentials.
-
Host org must be entitled to create trial orgs, and the invoking user must have sufficient access on it. The skill does
not run a separate permission check — the
call (Step 1) is the definitive gate, and the same API enforcement is what a check would rely on. If the org is not entitled, the
entity is not exposed and the create fails with a non-zero
and
/
of
("The requested resource does not exist") or
("sObject type 'SignupRequest' is not supported"). This case is handled in Step 1's error table.
When it happens,
stop (it is not retryable from the CLI) and report to the user: (a) the
raw CLI error as-is — the exact
/
and
the CLI returned, verbatim — and (b) that they should
reach out to Salesforce support to get the org enabled for trial-org creation, then try again.
Do NOT diagnose or name which permission is missing — just surface the raw error and point them to support.
Required Inputs — collect from the invoking user before any create
Prompt the user for these and do NOT proceed until all are provided. Do not invent values. Ask for each one; if the user is unsure about a field, guide them using the "If the user is unsure" column before moving on.
Always required (5) — these are
on the
entity (
is listed here for prompting convenience but is
optional):
| Input | Notes | If the user is unsure |
|---|
| Admin user's last name. Max 80 chars. | Any surname for the new org's admin user; it's just the admin contact name, use theirs. |
| Optional. Admin user's first name. Ask for it, but proceed without it if the user doesn't provide one. | Optional — leave blank if unsure; only is required for the admin user. |
| Admin login username. Must be email-format and globally unique across all Salesforce orgs. Max 80 chars. Lowercased on save. | It does not have to be a real inbox — it just has to look like an email and be unique. Suggest a pattern like admin@<company>-<something-unique>.com
. If it collides, you'll get a duplicate-username error on create; pick another. |
| Admin user's real email address (welcome/login mail goes here). | This one must be a working inbox they can access — unlike , it should be a real address. |
| Company / org name. Max 80 chars. | The organization name to show in the trial org; any descriptive name is fine. |
| ISO country code, max 3 chars, e.g. , , , . Validated at runtime against allowed codes (embargoed/invalid codes are rejected). | Use the 2-letter ISO code for their country (e.g. for United States, for United Kingdom). Not a free-text country name. |
Exactly one of (required, pick one — NOT both):
| Input | Notes | If the user is unsure |
|---|
| Trialforce template ID (key prefix , 15 chars) — defines the trial org's product/content. | Use a template when they want a specific pre-built product/content set. To find available templates, query the host org: sf data query -o <HOST_ORG> -q "SELECT Id, TemplateName FROM TrialforceTemplate" --json
. If they just want a plain trial org, use instead. |
| Org edition for a generic (non-template) trial. Generic values: , , , (also , ). Partner/Trialforce editions are perm-gated. | If they just want "a dev org to try things," use . Partner editions (, ) only work if the host org has partner/TMC perms — using one without the perm returns a error. |
Ask the user to choose
either a
or an
, not both:
- Neither supplied → stop and ask. A create with no template and no edition fails validation with (
ApiErrorCodes.INVALID_SIGNUP_OPTION
).
- Both supplied → ask them to pick one; send only the chosen field. Combining them fails with . Neither may be combined with clone/source-org fields either.
This skill intentionally scopes user-collected input to the fields above (the 5 always-required plus optional
). Do
not prompt for or surface other fields. The
entity supports additional optional and perm-gated fields (
,
,
,
, the OAuth-return pair, etc.); these are
out of scope here and left to server defaults. They are documented in
references/signup_request_fields.md
for reference only — do not send them from this skill.
Step 1 — Create the SignupRequest
Invoke the create script with the collected inputs. It enforces the "exactly one of
/
" rule, assembles and quotes the
payload, runs the insert, and prints the assigned
id on success. Reference the script by its
absolute path from the skill directory (
) — never
, which resolves against the user's working directory.
With a template:
bash
SR_ID=$(bash "<skill_dir>/scripts/create_signup_request.sh" \
--target-org <HOST_ORG> \
--last-name <LAST_NAME> --email <EMAIL> --username <UNIQUE_USERNAME> \
--company "<COMPANY>" --country <ISO> --template-id 0TT... \
--output-dir force-app/main/adk-eval-output)
With an edition (generic trial, no template): replace
with
(or
, etc.). Add
only if the user supplied it — no other optional fields are sent by this skill. The script rejects supplying both
and
, or neither.
Pass
(use
force-app/main/adk-eval-output
when it exists) so that if the create is
rejected, the script still writes
<output-dir>/signup-request-result.json
capturing the create-rejected outcome and the raw error verbatim — the run's output artifact even when no org is created. On success this write is done by the Step 2 read-back instead.
On success the script prints the
SignupRequest id (capture it as
). Pass
instead to get the raw
create envelope, which wraps a
handle, not the org:
json
{ "status": 0, "result": { "id": "0SRxx0000000000", "success": true, "errors": [] } }
Handle create errors (synchronous field validation). On a rejected create the script exits non-zero and prints the
raw CLI error (
/
+
) to stderr — surface it verbatim and act per the table below. When
was given, the script also writes the create-rejected artifact (
{ "outcome": "create-rejected", "error": {…}, "CreatedOrgId": null, "Status": null }
) to
<output-dir>/signup-request-result.json
; do not hand-author this file. This is field validation, returned
immediately — distinct from the async
in Step 2. These are
server-side rejections (the record reaches the org and the platform rejects it at insert time) — do not describe them to the user as "client-side"; the CLI does not validate email/country format, picklist values, or field length locally. Do NOT poll a create that failed, and do NOT proceed to Step 2.
| Failure | Meaning → what to tell the user |
|---|
| / | Neither nor was sent — ask for one and retry. |
| Both and were sent — drop one and retry. |
| / | A partner/TSO edition was requested but the host org lacks the perm — use a generic edition (, etc.) or get the perm. |
| duplicate / invalid () | is not email-format or not globally unique — ask for a different one and retry. |
| is not a valid/allowed ISO code — fix and retry. |
INVALID_OR_NULL_FOR_RESTRICTED_PICKLIST
| is not an accepted value for the host org's restricted picklist (e.g. ) — pick a valid generic edition and retry. |
| A field value exceeds its max length (the message names the field + , e.g. over 80) — shorten it and retry. |
| / invalid subdomain | Chosen is taken or invalid — pick another. |
| ("The requested resource does not exist") or ("sObject type 'SignupRequest' is not supported") | The entity is not exposed → the org is not entitled to create trial orgs. Not retryable. Stop and surface the raw CLI error as-is (/ + ), then tell the user to reach out to Salesforce support to get the org enabled. Do not name or diagnose the missing permission. |
INSUFFICIENT_ACCESS_OR_READONLY
| The entity is exposed but the user lacks the access to create the record — a user-permission problem, distinct from the org-entitlement failure above. Fix the user's permissions and retry. |
For the full catalog and prefixes →
references/error_codes.md
.
- For the full required/optional field list, types, and perm-gated fields → load
references/signup_request_fields.md
.
- Do NOT send together with clone/source-org fields ( error). Do NOT request a partner/Trialforce edition without the host org's partner/TMC perm.
Step 2 — Read the request once the org id is available (creation is async)
Provisioning happens asynchronously after the insert, so re-read the record to pick up the assigned org id. Invoke the read script — it applies a fixed, bounded read-back policy internally (stopping as soon as
is populated or the status is terminal, and never polling indefinitely), prints the record as JSON, and writes the output artifact when
is given. Then act on the script's
exit code (below) — the retry count and delay are the script's own deterministic logic; you do not re-implement or re-count them in prose. If the org id is not yet available, the script exits
so you can hand the request id back to the user:
bash
bash "<skill_dir>/scripts/get_signup_request.sh" \
--target-org <HOST_ORG> --id "$SR_ID" [--output-dir <DIR>]
The script prints the
record (the
envelope's
) as JSON. Read these fields from it:
- — the new trial org id (, 15 chars). Populated as soon as the org is allocated (often while is still ); this is the script's stop signal.
- — instance hosting the new org (target follow-up calls here).
- — the admin login username on the record. Report the value read back from the record, not the raw input — it is lowercased on save, so the stored value is the accurate one to hand the user.
- — lifecycle → → | (match case-insensitively).
- — present only if was set on create (perm-gated).
- — present only if + were set.
- — populated only when . This is the async provisioning error (distinct from the synchronous create-time validation in Step 1), prefixed:
- org creation error · signup data error · template error (e.g. = template not found) · org-shape error · version-selection error · / fatal/should-never-happen.
Act on the script's exit code.
The org id lookup runs first — do not report anything to the user until the script returns. Report only what is actually on the record; never invent or relabel the
:
- — is populated. Proceed to Step 3 and report the org id together with the record's real and the .
- — no org id available yet (not an error). The org has not been allocated. The script has already exhausted its bounded read-back — do NOT re-invoke it in a loop to keep polling. Report the SignupRequest id (), the record's exactly as returned, and the ; tell the user the org id is not available yet, and let them re-check later (see Step 3).
- — (stop and report the and its prefix meaning via
references/error_codes.md
; the org was not created), or the read itself failed (auth expired, id not found) — surface the raw CLI error and stop.
Step 3 — Report details and hand off status checks
Report
after the Step 2 lookup returns — not before. When the user asks to create (or re-check) the org, run the read-back first and wait for it, then report all available details to the user in one go. Report
only what is on the record; never fabricate or relabel a value — especially
, which must be the exact string the response object carries:
- SignupRequest id — the request id (from Step 1). Always report this — it is the handle the user (or a later check) uses to look the request up again, and it is the primary thing to hand back if the org id is not yet available.
- CreatedOrgId — the new trial org id (), when populated. If the read returned exit (org id not yet available), say so plainly: the request was accepted and the org is still being provisioned; there is no org id to share yet.
- CreatedOrgInstance — the instance hosting it, if present.
- Username — the admin login username as stored on the record (lowercased on save). Always report this — it is what the user logs in with once the org is ready.
- Status — the value exactly as it appears on the record. Echo whatever string the response carries; do not map, translate, infer, or pick from a fixed list. A still-pending status is normal at this point — provisioning finishes in the background.
- LoginUrl / AuthCode — only if present.
Also tell the user that
login details for the new org arrive by email — once provisioning completes, a welcome/login email is sent to the
address, so they should watch that inbox to finish logging in. (This is why
must be a real, accessible address.)
Then suggest, in plain language, how the user can re-check status later — provisioning may still be completing. Tell them they can just ask (the skill re-reads the record), for example:
- "Check the status of my trial org " (by the org id), or
- "Check the status of signup request " (by the SignupRequest id).
Both resolve to a re-run of the Step 2 read script against the same
record (the org id is looked up on that record). A later check should report the current
exactly as returned on the record. If the record comes back with an error status, also report the
and its prefix meaning via
references/error_codes.md
.
Output artifact. Write the current
record as the run's output artifact by passing
to the Step 2 read script — the script writes
<output-dir>/signup-request-result.json
(creating the directory if needed). Use
force-app/main/adk-eval-output
as the output directory when it exists:
bash
bash "<skill_dir>/scripts/get_signup_request.sh" \
--target-org <HOST_ORG> --id "$SR_ID" \
--output-dir force-app/main/adk-eval-output
This is the run's defined output — do not ask permission before writing it.
Reference File Index
| File | When to read |
|---|
references/signup_request_fields.md
| Full field reference — required/optional fields, types, defaults, perm-gated fields, editions |
references/error_codes.md
| Interpreting prefixes and common validation failures |
Example Files
Load these only when you need to see the concrete shape of a payload or response — they are illustrative samples with placeholder ids (
,
), not values to send.
| File | When to read |
|---|
examples/create_request.json
| When assembling the create — to confirm the field names/shape of the create payload |
examples/success_response.json
| When interpreting a successful create + read-back — shows the record once is populated |
examples/error_response.json
| When interpreting a rejected create — shows the shape of common validation/error responses |