Driving Detect Desk from code

Everything the web app does is available over HTTP. The base URL is https://api.skillsafe.ai/v1/app-api, every request carries Authorization: Bearer <token>, and every response is the same envelope.

The `task` field comes first

This app has five lanes behind one endpoint. Every run body must carry a task field naming the lane - it is what the system prompt routes on. Send the wrong one and you get a valid package of the wrong kind; omit it and the model picks the closest lane and tells you which it chose.

One more shape trap: the run body is the input object. Do not wrap it in an {"input": ...} envelope - that returns 200 while hiding task from the model, which is the most confusing way this API can fail.

task Lane Fields Sections returned

plan Design the measurement before you run it brief, known Summary, The Plan, The Numbers, Reasoning, Next Step

read Is this difference real, and what could this run have seen sheet, worry Summary, Verdict, Findings, Corrected Plan, Next Step

power The detection floor: what this suite can and cannot see sheet Summary, What Each Pair Could Detect, What It Would Take, What This Run Actually Tested, Next Step

tail Percentiles, outliers, and which statistic to quote sheet Summary, Every Percentile Asked For, What One Point Is Doing, Which Statistic To Quote, Next Step

decide Decide what changes: more runs, a quieter machine, or the question sheet, fixed Summary, More Runs Fixes, Only A Different Measurement Fixes, Nothing Fixes, Next Step

task	Lane	Fields	Sections returned
`plan`	Design the measurement before you run it	`brief, known`	Summary, The Plan, The Numbers, Reasoning, Next Step
`read`	Is this difference real, and what could this run have seen	`sheet, worry`	Summary, Verdict, Findings, Corrected Plan, Next Step
`power`	The detection floor: what this suite can and cannot see	`sheet`	Summary, What Each Pair Could Detect, What It Would Take, What This Run Actually Tested, Next Step
`tail`	Percentiles, outliers, and which statistic to quote	`sheet`	Summary, Every Percentile Asked For, What One Point Is Doing, Which Statistic To Quote, Next Step
`decide`	Decide what changes: more runs, a quieter machine, or the question	`sheet, fixed`	Summary, More Runs Fixes, Only A Different Measurement Fixes, Nothing Fixes, Next Step

Only task and the lane's own required fields are mandatory: sheet on read, power, tail and decide; brief on plan. Every field is a string - there are no number fields on this app. sheet is the results sheet itself: a header of KEY: value lines and one block - RUNS: for measurements, SUMMARY: for a table that already has n, mean and sd, or TASKS: for pass counts.

The first variant is the BASELINE and every other is compared against it. That is why the order is part of the grammar: which variant is the reference decides every delta, every interval and which way round the percentages read.

A SUMMARY: block is a first-class input, not a degraded one. Welch's t needs only n, mean and sd from each side, which is exactly what a benchmark harness's own summary table prints - so pasting it gives the same test, the same interval and the same detection floor. What it cannot give is the tail, the median, or what a single extreme run is doing, all of which need the runs themselves. A long RUNS: list may wrap onto indented continuation lines that start with a number.

ALPHA is the false-alarm rate per comparison and decides every verdict; assumed 0.05. POWER is the chance of noticing a real effect of the size you asked about; assumed 0.8, which means accepting a one-in-five chance of missing it. TARGET is the regression size you need to catch - state it and the response will say whether the run count can resolve it. PERCENTILE chooses which tail is reported; assumed 0.95, and a warning about the tail is only raised when you asked for one. UNIT is echoed on every figure and never converted, and BETTER says which direction is an improvement; assumed lower.

Anything the reader cannot place is listed as a problem rather than skipped, and so is a variant named twice. A variant that quietly vanished would change the comparison count, the family-wise rate and which variant is the baseline - so a swallowed line moves every figure in the response at once.

A measurement is reported in its declared unit, a relative change as a signed percentage, a p-value to four places (or <0.0001), a t to two, a share as a percentage and a run count as a whole number. The engine computes Welch's t with Satterthwaite degrees of freedom, the exact p from the regularised incomplete beta, the minimum detectable effect, binomial rank intervals and Wilson intervals - all to double precision. It ran nothing and timed nothing.

Add $model to any body to choose the model for that run: gpt-5.6-luna, gpt-5.6-terra (the default) or gpt-5.6-sol. Luna caps output at 4,096 tokens and will fail the read, power and tail lanes rather than shorten them - a findings table, a corrected plan, or three tables with a row per variant or per pair, is several thousand characters before the reasoning starts.

The response envelope

Success and failure have the same outer shape, so one check covers both.

{
  "ok": true,
  "data": {
    "...": "the result"
  }
}

{
  "ok": false,
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "seconds should be number, got string",
    "details": {}
  }
}

HTTP error.code What it means

400 VALIDATION_ERROR The body was not a JSON object, or a declared field had the wrong type. A number field sent as a string is the usual cause.

401 UNAUTHORIZED No token, or a token that has expired or been revoked. Mint a new one.

402 INSUFFICIENT_CREDITS The balance is below the run's minimum. Call /estimate first and compare hold_credits against /me.

404 NOT_FOUND Wrong path, or a job id that does not belong to this token.

409 CONFLICT An Idempotency-Key replay whose body differs from the original request.

429 RATE_LIMITED Too many requests. Back off; do not tight-loop.

503 UPSTREAM_UNAVAILABLE The model provider is unavailable. Retry with backoff.

HTTP	error.code	What it means
`400`	`VALIDATION_ERROR`	The body was not a JSON object, or a declared field had the wrong type. A number field sent as a string is the usual cause.
`401`	`UNAUTHORIZED`	No token, or a token that has expired or been revoked. Mint a new one.
`402`	`INSUFFICIENT_CREDITS`	The balance is below the run's minimum. Call /estimate first and compare hold_credits against /me.
`404`	`NOT_FOUND`	Wrong path, or a job id that does not belong to this token.
`409`	`CONFLICT`	An Idempotency-Key replay whose body differs from the original request.
`429`	`RATE_LIMITED`	Too many requests. Back off; do not tight-loop.
`503`	`UPSTREAM_UNAVAILABLE`	The model provider is unavailable. Retry with backoff.

1. Get a token

Open /tokens.html in a browser and copy the token this app already holds - no developer console needed. A guest token is minted automatically and is enough for /me and /estimate; writing a package is metered and needs a personal token, which comes from signing in on that page.

Keep it in an environment variable rather than in source:

export SKILLSAFE_TOKEN="YOUR_TOKEN"

2. Check the session and the balance

GET /me is free. It returns only three fields: subject_type, subject_id and credits. Signed-in means subject_type == "user" - there is no username or email to test.

curl -sS -X GET "https://api.skillsafe.ai/v1/app-api/me" \
  -H "Authorization: Bearer $SKILLSAFE_TOKEN"

import json, urllib.request

TOKEN = "YOUR_TOKEN"   # from /tokens.html
BASE  = "https://api.skillsafe.ai/v1/app-api"

def call(method, path, payload=None, headers=None):
    data = json.dumps(payload).encode() if payload is not None else None
    req = urllib.request.Request(BASE + path, data=data, method=method)
    req.add_header("Authorization", "Bearer " + TOKEN)
    req.add_header("Content-Type", "application/json")
    req.add_header("Accept", "application/json")
    # A plain urllib request with no User-Agent is refused with 403.
    req.add_header("User-Agent", "detect-desk-client/1.0")
    for k, v in (headers or {}).items():
        req.add_header(k, v)
    with urllib.request.urlopen(req) as r:
        return json.loads(r.read())

out = call("GET", "/me")
print(json.dumps(out["data"], indent=2))

const TOKEN = "YOUR_TOKEN";   // from /tokens.html
const BASE  = "https://api.skillsafe.ai/v1/app-api";

async function call(method, path, payload) {
  const res = await fetch(BASE + path, {
    method,
    headers: {
      "Authorization": `Bearer ${TOKEN}`,
      "Content-Type": "application/json"
    },
    body: payload === undefined ? undefined : JSON.stringify(payload)
  });
  const json = await res.json();
  if (!res.ok || !json.ok) throw new Error(json.error?.message || res.statusText);
  return json.data;
}

const data = await call("GET", "/me");
console.log(data);

package main

import (
	"bytes"
	"fmt"
	"io"
	"net/http"
)

const base = "https://api.skillsafe.ai/v1/app-api"
const token = "YOUR_TOKEN" // from /tokens.html

func main() {
	req, _ := http.NewRequest("GET", base+"/me", nil)
	req.Header.Set("Authorization", "Bearer "+token)
	req.Header.Set("Content-Type", "application/json")
	res, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer res.Body.Close()
	out, _ := io.ReadAll(res.Body)
	fmt.Println(string(out))
}

import java.net.URI;
import java.net.http.*;

public class Main {
    static final String BASE  = "https://api.skillsafe.ai/v1/app-api";
    static final String TOKEN = "YOUR_TOKEN"; // from /tokens.html

    public static void main(String[] args) throws Exception {
        HttpRequest req = HttpRequest.newBuilder()
            .uri(URI.create(BASE + "/me"))
            .header("Authorization", "Bearer " + TOKEN)
            .header("Content-Type", "application/json")
            .method("GET", HttpRequest.BodyPublishers.noBody())
            .build();
        HttpResponse<String> res = HttpClient.newHttpClient()
            .send(req, HttpResponse.BodyHandlers.ofString());
        System.out.println(res.body());
    }
}

require 'net/http'
require 'json'
require 'uri'

BASE  = "https://api.skillsafe.ai/v1/app-api"
TOKEN = "YOUR_TOKEN" # from /tokens.html

uri = URI("#{BASE}/me")
req = Net::HTTP::Get.new(uri)
req["Authorization"] = "Bearer #{TOKEN}"
req["Content-Type"] = "application/json"
res = Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) { |h| h.request(req) }
puts JSON.pretty_generate(JSON.parse(res.body)["data"])

<?php
$base  = "https://api.skillsafe.ai/v1/app-api";
$token = "YOUR_TOKEN"; // from /tokens.html

$opts = ["http" => [
    "method"  => "GET",
    "header"  => "Authorization: Bearer $token\r\nContent-Type: application/json\r\n",
]];
$res  = file_get_contents($base . "/me", false, stream_context_create($opts));
$json = json_decode($res, true);
print_r($json["data"]);

using System.Net.Http;
using System.Text;

const string Base  = "https://api.skillsafe.ai/v1/app-api";
const string Token = "YOUR_TOKEN"; // from /tokens.html

var http = new HttpClient();
http.DefaultRequestHeaders.Add("Authorization", $"Bearer {Token}");

var res = await http.SendAsync(new HttpRequestMessage(new HttpMethod("GET"), Base + "/me")
{
    Content = null
});
Console.WriteLine(await res.Content.ReadAsStringAsync());

3. Price the run before making it

POST /estimate costs nothing, creates no job, and returns the worst-case cost. Compare hold_credits against the balance from step 2 before you submit: a 402 after the fact is avoidable. hold_credits is a reservation priced at the full output cap - the actual charge is usually far lower.

It also echoes model, model_alias and markup_bps, which is the authoritative check that a run is bound to the model you think it is. Estimate each lane separately: their prompts and caps differ, so their holds do.

curl -sS -X POST "https://api.skillsafe.ai/v1/app-api/estimate" \
  -H "Authorization: Bearer $SKILLSAFE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}'

import json, urllib.request

TOKEN = "YOUR_TOKEN"   # from /tokens.html
BASE  = "https://api.skillsafe.ai/v1/app-api"

def call(method, path, payload=None, headers=None):
    data = json.dumps(payload).encode() if payload is not None else None
    req = urllib.request.Request(BASE + path, data=data, method=method)
    req.add_header("Authorization", "Bearer " + TOKEN)
    req.add_header("Content-Type", "application/json")
    req.add_header("Accept", "application/json")
    # A plain urllib request with no User-Agent is refused with 403.
    req.add_header("User-Agent", "detect-desk-client/1.0")
    for k, v in (headers or {}).items():
        req.add_header(k, v)
    with urllib.request.urlopen(req) as r:
        return json.loads(r.read())

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
out = call("POST", "/estimate", payload)
print(out["data"]["hold_credits"], out["data"]["model"])

const TOKEN = "YOUR_TOKEN";   // from /tokens.html
const BASE  = "https://api.skillsafe.ai/v1/app-api";

async function call(method, path, payload) {
  const res = await fetch(BASE + path, {
    method,
    headers: {
      "Authorization": `Bearer ${TOKEN}`,
      "Content-Type": "application/json"
    },
    body: payload === undefined ? undefined : JSON.stringify(payload)
  });
  const json = await res.json();
  if (!res.ok || !json.ok) throw new Error(json.error?.message || res.statusText);
  return json.data;
}

const payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
};
const data = await call("POST", "/estimate", payload);
console.log(data.hold_credits, data.model, data.model_alias);

package main

import (
	"bytes"
	"fmt"
	"io"
	"net/http"
)

const base = "https://api.skillsafe.ai/v1/app-api"
const token = "YOUR_TOKEN" // from /tokens.html

func main() {
	payload := []byte(`{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}`)
	req, _ := http.NewRequest("POST", base+"/estimate", bytes.NewReader(payload))
	req.Header.Set("Authorization", "Bearer "+token)
	req.Header.Set("Content-Type", "application/json")
	res, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer res.Body.Close()
	out, _ := io.ReadAll(res.Body)
	fmt.Println(string(out))
}

import java.net.URI;
import java.net.http.*;

public class Main {
    static final String BASE  = "https://api.skillsafe.ai/v1/app-api";
    static final String TOKEN = "YOUR_TOKEN"; // from /tokens.html

    public static void main(String[] args) throws Exception {
        String payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
        HttpRequest req = HttpRequest.newBuilder()
            .uri(URI.create(BASE + "/estimate"))
            .header("Authorization", "Bearer " + TOKEN)
            .header("Content-Type", "application/json")
            .method("POST", HttpRequest.BodyPublishers.ofString(payload))
            .build();
        HttpResponse<String> res = HttpClient.newHttpClient()
            .send(req, HttpResponse.BodyHandlers.ofString());
        System.out.println(res.body());
    }
}

require 'net/http'
require 'json'
require 'uri'

BASE  = "https://api.skillsafe.ai/v1/app-api"
TOKEN = "YOUR_TOKEN" # from /tokens.html

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}

uri = URI("#{BASE}/estimate")
req = Net::HTTP::Post.new(uri)
req["Authorization"] = "Bearer #{TOKEN}"
req["Content-Type"] = "application/json"
req.body = JSON.dump(payload)
res = Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) { |h| h.request(req) }
puts JSON.pretty_generate(JSON.parse(res.body)["data"])

<?php
$base  = "https://api.skillsafe.ai/v1/app-api";
$token = "YOUR_TOKEN"; // from /tokens.html

$payload = json_encode([
  "task" => "check",
  "sheet" => "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry" => "it graded fine last time and this one will not come clean",
  "rules" => "<the working rules for this lane, sent by the app>"
]);

$opts = ["http" => [
    "method"  => "POST",
    "header"  => "Authorization: Bearer $token\r\nContent-Type: application/json\r\n",
    "content" => $payload,
]];
$res  = file_get_contents($base . "/estimate", false, stream_context_create($opts));
$json = json_decode($res, true);
print_r($json["data"]);

using System.Net.Http;
using System.Text;

const string Base  = "https://api.skillsafe.ai/v1/app-api";
const string Token = "YOUR_TOKEN"; // from /tokens.html

var http = new HttpClient();
http.DefaultRequestHeaders.Add("Authorization", $"Bearer {Token}");

var payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
var content = new StringContent(payload, Encoding.UTF8, "application/json");
var res = await http.SendAsync(new HttpRequestMessage(new HttpMethod("POST"), Base + "/estimate")
{
    Content = content
});
Console.WriteLine(await res.Content.ReadAsStringAsync());

4. Write a package

POST /run submits the job. Always send an Idempotency-Key: a network blip that replays the same request must not bill twice. A replay with the same key returns the stored result and is not charged again; a replay with the same key but a different body is a 409.

The response carries output.output (the Markdown package), charged_credits and truncated. If truncated is true the balance sat between min_credits and hold_credits and the output was cut short - render what arrived and say so rather than presenting it as complete.

curl -sS -X POST "https://api.skillsafe.ai/v1/app-api/run" \
  -H "Authorization: Bearer $SKILLSAFE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}'

import json, urllib.request

TOKEN = "YOUR_TOKEN"   # from /tokens.html
BASE  = "https://api.skillsafe.ai/v1/app-api"

def call(method, path, payload=None, headers=None):
    data = json.dumps(payload).encode() if payload is not None else None
    req = urllib.request.Request(BASE + path, data=data, method=method)
    req.add_header("Authorization", "Bearer " + TOKEN)
    req.add_header("Content-Type", "application/json")
    req.add_header("Accept", "application/json")
    # A plain urllib request with no User-Agent is refused with 403.
    req.add_header("User-Agent", "detect-desk-client/1.0")
    for k, v in (headers or {}).items():
        req.add_header(k, v)
    with urllib.request.urlopen(req) as r:
        return json.loads(r.read())

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
out = call("POST", "/run", payload)
print(out["data"]["output"]["output"])

const TOKEN = "YOUR_TOKEN";   // from /tokens.html
const BASE  = "https://api.skillsafe.ai/v1/app-api";

async function call(method, path, payload) {
  const res = await fetch(BASE + path, {
    method,
    headers: {
      "Authorization": `Bearer ${TOKEN}`,
      "Content-Type": "application/json"
    },
    body: payload === undefined ? undefined : JSON.stringify(payload)
  });
  const json = await res.json();
  if (!res.ok || !json.ok) throw new Error(json.error?.message || res.statusText);
  return json.data;
}

const payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
};
const data = await call("POST", "/run", payload);
console.log(data.output.output);

package main

import (
	"bytes"
	"fmt"
	"io"
	"net/http"
)

const base = "https://api.skillsafe.ai/v1/app-api"
const token = "YOUR_TOKEN" // from /tokens.html

func main() {
	payload := []byte(`{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}`)
	req, _ := http.NewRequest("POST", base+"/run", bytes.NewReader(payload))
	req.Header.Set("Authorization", "Bearer "+token)
	req.Header.Set("Content-Type", "application/json")
	res, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer res.Body.Close()
	out, _ := io.ReadAll(res.Body)
	fmt.Println(string(out))
}

import java.net.URI;
import java.net.http.*;

public class Main {
    static final String BASE  = "https://api.skillsafe.ai/v1/app-api";
    static final String TOKEN = "YOUR_TOKEN"; // from /tokens.html

    public static void main(String[] args) throws Exception {
        String payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
        HttpRequest req = HttpRequest.newBuilder()
            .uri(URI.create(BASE + "/run"))
            .header("Authorization", "Bearer " + TOKEN)
            .header("Content-Type", "application/json")
            .method("POST", HttpRequest.BodyPublishers.ofString(payload))
            .build();
        HttpResponse<String> res = HttpClient.newHttpClient()
            .send(req, HttpResponse.BodyHandlers.ofString());
        System.out.println(res.body());
    }
}

require 'net/http'
require 'json'
require 'uri'

BASE  = "https://api.skillsafe.ai/v1/app-api"
TOKEN = "YOUR_TOKEN" # from /tokens.html

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}

uri = URI("#{BASE}/run")
req = Net::HTTP::Post.new(uri)
req["Authorization"] = "Bearer #{TOKEN}"
req["Content-Type"] = "application/json"
req.body = JSON.dump(payload)
res = Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) { |h| h.request(req) }
puts JSON.pretty_generate(JSON.parse(res.body)["data"])

<?php
$base  = "https://api.skillsafe.ai/v1/app-api";
$token = "YOUR_TOKEN"; // from /tokens.html

$payload = json_encode([
  "task" => "check",
  "sheet" => "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry" => "it graded fine last time and this one will not come clean",
  "rules" => "<the working rules for this lane, sent by the app>"
]);

$opts = ["http" => [
    "method"  => "POST",
    "header"  => "Authorization: Bearer $token\r\nContent-Type: application/json\r\n",
    "content" => $payload,
]];
$res  = file_get_contents($base . "/run", false, stream_context_create($opts));
$json = json_decode($res, true);
print_r($json["data"]);

using System.Net.Http;
using System.Text;

const string Base  = "https://api.skillsafe.ai/v1/app-api";
const string Token = "YOUR_TOKEN"; // from /tokens.html

var http = new HttpClient();
http.DefaultRequestHeaders.Add("Authorization", $"Bearer {Token}");

var payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
var content = new StringContent(payload, Encoding.UTF8, "application/json");
var res = await http.SendAsync(new HttpRequestMessage(new HttpMethod("POST"), Base + "/run")
{
    Content = content
});
Console.WriteLine(await res.Content.ReadAsStringAsync());

5. Stream a run

POST /run-stream is the same call with a text/event-stream response. Worth knowing before you build on it: from a server or from cURL you get event: delta frames carrying the output token by token; from a browser you get event: tick heartbeats and then one event: done with the whole output. Handle both, and treat ticks as liveness rather than progress.

Frame types are job (the job id), delta ({"text": "..."}), tick ({"t": seconds}), done, and error. An idempotent replay returns plain JSON with no stream at all, so check the content type before you start reading frames.

curl -sS -N -X POST "https://api.skillsafe.ai/v1/app-api/run-stream" \
  -H "Authorization: Bearer $SKILLSAFE_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -H "Idempotency-Key: cbd-$(date +%s)" \
  -d '{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}'

import json, urllib.request

TOKEN = "YOUR_TOKEN"
BASE  = "https://api.skillsafe.ai/v1/app-api"

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}

req = urllib.request.Request(BASE + "/run-stream", data=json.dumps(payload).encode(), method="POST")
req.add_header("Authorization", "Bearer " + TOKEN)
req.add_header("Content-Type", "application/json")
req.add_header("Accept", "text/event-stream")
req.add_header("Idempotency-Key", "cbd-demo-1")
req.add_header("User-Agent", "detect-desk-client/1.0")

event, data = "message", ""
with urllib.request.urlopen(req) as r:
    for raw in r:
        line = raw.decode().rstrip("\n")
        if line.startswith("event:"):
            event = line[6:].strip()
        elif line.startswith("data:"):
            data += line[5:].strip()
        elif line == "":
            if data:
                frame = json.loads(data)
                if event == "delta":
                    print(frame.get("text", ""), end="")
                elif event == "done":
                    print("\n--- charged:", frame.get("charged_credits"))
            event, data = "message", ""

const TOKEN = "YOUR_TOKEN";
const BASE  = "https://api.skillsafe.ai/v1/app-api";

const payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
};

const res = await fetch(BASE + "/run-stream", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${TOKEN}`,
    "Content-Type": "application/json",
    "Idempotency-Key": "cbd-demo-1"
  },
  body: JSON.stringify(payload)
});

// An idempotent replay comes back as plain JSON with no stream at all.
if (!(res.headers.get("content-type") || "").includes("text/event-stream")) {
  const json = await res.json();
  console.log(json.data.output.output);
} else {
  const reader = res.body.getReader();
  const dec = new TextDecoder();
  let buf = "";
  for (;;) {
    const { done, value } = await reader.read();
    if (done) break;
    buf += dec.decode(value, { stream: true });
    let i;
    while ((i = buf.indexOf("\n\n")) >= 0) {
      const frame = buf.slice(0, i);
      buf = buf.slice(i + 2);
      let name = "message", data = "";
      for (const line of frame.split("\n")) {
        if (line.startsWith("event:")) name = line.slice(6).trim();
        else if (line.startsWith("data:")) data += line.slice(5).trim();
      }
      if (!data) continue;
      const p = JSON.parse(data);
      if (name === "delta") process.stdout.write(p.text || "");
      if (name === "tick") console.error("still writing:", p.t + "s");
      if (name === "done") console.log("\ncharged:", p.charged_credits);
    }
  }
}

package main

import (
	"bytes"
	"fmt"
	"io"
	"net/http"
)

const base = "https://api.skillsafe.ai/v1/app-api"
const token = "YOUR_TOKEN" // from /tokens.html

func main() {
	payload := []byte(`{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}`)
	req, _ := http.NewRequest("POST", base+"/run-stream", bytes.NewReader(payload))
	req.Header.Set("Authorization", "Bearer "+token)
	req.Header.Set("Content-Type", "application/json")
	res, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer res.Body.Close()
	out, _ := io.ReadAll(res.Body)
	fmt.Println(string(out))
}

import java.net.URI;
import java.net.http.*;

public class Main {
    static final String BASE  = "https://api.skillsafe.ai/v1/app-api";
    static final String TOKEN = "YOUR_TOKEN"; // from /tokens.html

    public static void main(String[] args) throws Exception {
        String payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
        HttpRequest req = HttpRequest.newBuilder()
            .uri(URI.create(BASE + "/run-stream"))
            .header("Authorization", "Bearer " + TOKEN)
            .header("Content-Type", "application/json")
            .method("POST", HttpRequest.BodyPublishers.ofString(payload))
            .build();
        HttpResponse<String> res = HttpClient.newHttpClient()
            .send(req, HttpResponse.BodyHandlers.ofString());
        System.out.println(res.body());
    }
}

require 'net/http'
require 'json'
require 'uri'

BASE  = "https://api.skillsafe.ai/v1/app-api"
TOKEN = "YOUR_TOKEN" # from /tokens.html

payload = {
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}

uri = URI("#{BASE}/run-stream")
req = Net::HTTP::Post.new(uri)
req["Authorization"] = "Bearer #{TOKEN}"
req["Content-Type"] = "application/json"
req.body = JSON.dump(payload)
res = Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) { |h| h.request(req) }
puts JSON.pretty_generate(JSON.parse(res.body)["data"])

<?php
$base  = "https://api.skillsafe.ai/v1/app-api";
$token = "YOUR_TOKEN"; // from /tokens.html

$payload = json_encode([
  "task" => "check",
  "sheet" => "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry" => "it graded fine last time and this one will not come clean",
  "rules" => "<the working rules for this lane, sent by the app>"
]);

$opts = ["http" => [
    "method"  => "POST",
    "header"  => "Authorization: Bearer $token\r\nContent-Type: application/json\r\n",
    "content" => $payload,
]];
$res  = file_get_contents($base . "/run-stream", false, stream_context_create($opts));
$json = json_decode($res, true);
print_r($json["data"]);

using System.Net.Http;
using System.Text;

const string Base  = "https://api.skillsafe.ai/v1/app-api";
const string Token = "YOUR_TOKEN"; // from /tokens.html

var http = new HttpClient();
http.DefaultRequestHeaders.Add("Authorization", $"Bearer {Token}");

var payload = """
{
  "task": "audit",
  "sheet": "<UNIT, ALPHA and a RUNS, SUMMARY or TASKS block, baseline first; the grammar is in /llms.txt>",
  "worry": "it graded fine last time and this one will not come clean",
  "rules": "<the working rules for this lane, sent by the app>"
}
""";
var content = new StringContent(payload, Encoding.UTF8, "application/json");
var res = await http.SendAsync(new HttpRequestMessage(new HttpMethod("POST"), Base + "/run-stream")
{
    Content = content
});
Console.WriteLine(await res.Content.ReadAsStringAsync());

6. Read the result

output.output is Markdown in the envelope this app's system prompt guarantees: every section is a level-two heading spelled exactly as listed in the lane table above, in that order; tables are GitHub pipe tables with the declared columns; prompts are in fenced blocks opened with three backticks and the word text; checklists are - [x] lines.

So parsing is a split on /^## / - but do it fence-aware, because a prompt block can legitimately contain a line starting with ##. Count the sections you got against the ones the lane declares: a short list means the run was truncated, not that the contract changed.

def sections(md):
    out, name, buf, fence = {}, None, [], False
    for line in md.split("\n"):
        if line.lstrip().startswith("```"):
            fence = not fence
        if not fence and line.startswith("## "):
            if name:
                out[name] = "\n".join(buf).strip()
            name, buf = line[3:].strip(), []
            continue
        if name:
            buf.append(line)
    if name:
        out[name] = "\n".join(buf).strip()
    return out

The artifact most callers want is the fenced text block inside ## The Sheet or ## Corrected Sheet - that is a complete sheet in the grammar above, so it can be fed straight back into another lane with nothing carried alongside it. Every other section is prose and tables meant to be read.

Rate limits and good manners

/estimate and /me are free. Call them as much as you like, within reason.
A 429 means back off with a delay, not retry immediately.
Send an Idempotency-Key on every /run. Derive it from a hash of the body plus an attempt counter, so a retry of the same request reuses the key and a deliberate re-run gets a new one.
The engine in the web app is client-side only and has no endpoint. What it computes is published in full in llms.txt, and it is one idea and five consequences. A benchmark is a sample, so the difference between two variants is a random quantity and the headline percentage is one draw from it. Welch's t with Satterthwaite degrees of freedom tests it - deliberately not Student's, because two variants rarely share a variance and assuming they do is what turns a noisy candidate into a significant result - and its exact two-sided p comes from the regularised incomplete beta, I_df/(df+t²)(df/2, ½). The minimum detectable effect is (z_α/2 + z_power) × cv × √(2/n) and the run count is the same formula read backwards, n = 2 × ((z + z) × cv / δ)² — both depend on the noise, the run count and the two rates you chose, never on the result. The family-wise false-alarm rate across k comparisons is 1 − (1 − α)^k, and the response carries Holm's step-down thresholds, which hold the family at alpha, are uniformly more powerful than Bonferroni, and stay valid when the comparisons are dependent. A percentile is an order statistic: the count of samples below the true quantile is Binomial(n, p), so the rank interval is exact and assumes nothing about the distribution's shape. And a pass rate is a proportion, reported with a Wilson interval rather than a normal one because the normal interval degenerates to [100%, 100%] on a clean sweep. So you can compute every figure this app produces in your own pipeline without it. It ran nothing and timed nothing: these are the numbers you pasted, and whether they came from a fair comparison is the harder half that this page cannot see. The runs are assumed independent and comparable, which consecutive runs on one machine usually are not - interleaving the variants is what buys that. A p-value is not the chance the change did nothing, and not separable is not the same as the same. Every variant with its spread and interval, every comparison with its p, its interval AND the smallest effect it could have detected, the family-wise rate with every Holm threshold, and the exact rank interval of every percentile are all sent with each run as prescan, and the system prompt requires answering them.