[udate]HuggingFace access tken

Author: tsato-fuji

README.md

@@ -14,19 +14,6 @@ https://en-documents.research.global.fujitsu.com/fieldworkarena/
 
 ## Prerequisites
 
-### Set up OpenAI API Key
-
-The green agent uses LLM-as-a-judge for evaluation, which requires an OpenAI API key. You need to add this key as a GitHub secret:
-
-1. Go to your repository's **Settings** → **Secrets and variables** → **Actions**
-2. Click **New repository secret**
-3. Set the following:
-   - **Name**: `OPENAI_API_KEY`
-   - **Secret**: Your OpenAI API key (starts with `sk-`)
-4. Click **Add secret**
-
-**Note:** This API key will be used by the green agent to evaluate participant submissions using LLM-as-a-judge methodology.
-
 ### Request Access to Hugging Face Dataset
 
 This assessment requires access to the FieldWorkArena dataset hosted on Hugging Face. To request access:
@@ -43,6 +30,34 @@ This assessment requires access to the FieldWorkArena dataset hosted on Hugging
 
 **Note:** You must have an approved access token before running the benchmark tasks. Please note that access permission handling procedures may be subject to change.
 
+### Set up GitHub Secrets
+
+The benchmark requires two secrets to be configured in your GitHub repository:
+
+#### OPENAI_API_KEY
+The green agent uses LLM-as-a-judge for evaluation, which requires an OpenAI API key.
+
+1. Go to your repository's **Settings** → **Secrets and variables** → **Actions**
+2. Click **New repository secret**
+3. Set the following:
+   - **Name**: `OPENAI_API_KEY`
+   - **Secret**: Your OpenAI API key (starts with `sk-`)
+4. Click **Add secret**
+
+**Note:** This API key will be used by the green agent to evaluate participant submissions using LLM-as-a-judge methodology.
+
+#### HF_TOKEN
+The Hugging Face access token is required to access the FieldWorkArena dataset.
+
+1. Go to your repository's **Settings** → **Secrets and variables** → **Actions**
+2. Click **New repository secret**
+3. Set the following:
+   - **Name**: `HF_TOKEN`
+   - **Secret**: Your Hugging Face access token (obtained from the dataset access request)
+4. Click **Add secret**
+
+**Note:** This token must have the necessary permissions to access the FieldWorkArena dataset on Hugging Face.
+
 ## Scoring
 
 Participant agents are evaluated on their ability to accurately complete real-world field tasks. The evaluation uses LLM-as-a-judge methodology to assess:
@@ -66,15 +81,17 @@ Specifies the target category of tasks to run. Available options:
 - `"all"`: Runs all available task categories
 
 ### token
-Your FieldWorkArena-authenticated Hugging Face access token. This token is required to access the FieldWorkArena dataset. See the [Prerequisites](#request-access-to-hugging-face-dataset) section for instructions on obtaining this token.
+Your FieldWorkArena-authenticated Hugging Face access token. This token is required to access the FieldWorkArena dataset. 
 
-Example configuration in `scenario.toml`:
+In `scenario.toml`, configure it to use the `HF_TOKEN` environment variable from GitHub Secrets:
 ```toml
 [config]
 target = "factory"
-token = "hf_xxxxxxxxxxxxxxxxxxxxx"
+token = "${HF_TOKEN}"
 ```
 
+**Note:** The `${HF_TOKEN}` value will be automatically replaced with your Hugging Face access token from GitHub Secrets during GitHub Actions workflow execution. See the [Prerequisites](#set-up-github-secrets) section for instructions on setting up this secret.
+
 ## Requirements for participant agent
 
 Participant agents (Purple Agents) must meet the following requirements:

scenario.toml

@@ -10,4 +10,4 @@ env = {}
 
 [config]
 target = "custom"
-token=""
+token="${HF_TOKEN}"