customgpt-ai-rag:create-agent

create-agent

Create a CustomGPT.ai agent for a local folder and upload all eligible files into its knowledge base.

Rules

• NEVER proceed without an API key
• ALWAYS save .customgpt-meta.json before reporting success
• NEVER upload .git/, node_modules/, secrets (.env, .env.*), or binary files
• If .customgpt-meta.json already exists in the target folder, warn the user before overwriting — they may want /refresh-agent instead

---

Step 1 — Resolve API Key

Follow the lookup procedure in skills/_shared/api-key.md. Store the result as $API_KEY.

---

Step 2 — Determine the Folder

If the user specified a path, resolve it to an absolute path and verify it exists:

realpath "$USER_PATH"

If no path was given, use the Git repository root if available, otherwise $PWD:

git rev-parse --show-toplevel 2>/dev/null || echo "$PWD"

Store the result as $FOLDER. Tell the user: "Creating agent for: {FOLDER}"

Check whether .customgpt-meta.json already exists in $FOLDER. If it does, read it and ask:

"An agent already exists for this folder: '{agent_name}' (ID: {agent_id}). Run /refresh-agent to re-sync it, or type 'overwrite' to create a new agent and replace the meta file."

If they don't confirm overwrite, stop.

---

Step 3 — Create the Agent

curl -s --request POST \
  --url "https://app.customgpt.ai/api/v1/projects" \
  --header "Authorization: Bearer ${API_KEY}" \
  --form "project_name=$(basename $FOLDER)" \
  --form "file_data_retension=true"

Read the response. On success (HTTP 201), extract data.id as $AGENT_ID and data.project_name.

On failure, show the full response and stop. Common errors:

• HTTP 401 → invalid API key (see skills/_shared/api-key.md)
• HTTP 400 → check that project_name is not empty

---

Step 4 — Collect Eligible Files

find "$FOLDER" -type f \
  -not -path "*/.git/*" \
  -not -path "*/node_modules/*" \
  -not -path "*/__pycache__/*" \
  -not -path "*/.next/*" \
  -not -path "*/dist/*" \
  -not -path "*/build/*" \
  -not -path "*/.cache/*" \
  -not -path "*/vendor/*" \
  -not -path "*/coverage/*" \
  -not -path "*/.venv/*" \
  -not -path "*/venv/*" \
  -not -path "*/target/*" \
  -not -path "*/.turbo/*" \
  -not -path "*/.parcel-cache/*" \
  -not -name ".env" \
  -not -name ".env.*" \
  \( \
    -name "*.js" -o -name "*.ts" -o -name "*.tsx" -o -name "*.jsx" \
    -o -name "*.mjs" -o -name "*.cjs" \
    -o -name "*.py" -o -name "*.go" -o -name "*.rb" -o -name "*.java" \
    -o -name "*.cs" -o -name "*.cpp" -o -name "*.c" -o -name "*.h" -o -name "*.hpp" \
    -o -name "*.rs" -o -name "*.swift" -o -name "*.kt" -o -name "*.php" \
    -o -name "*.scala" -o -name "*.lua" -o -name "*.r" -o -name "*.R" \
    -o -name "*.sh" -o -name "*.bash" -o -name "*.zsh" -o -name "*.fish" \
    -o -name "*.html" -o -name "*.htm" -o -name "*.css" -o -name "*.scss" \
    -o -name "*.sass" -o -name "*.less" -o -name "*.svelte" -o -name "*.vue" \
    -o -name "*.sql" -o -name "*.graphql" -o -name "*.proto" \
    -o -name "*.json" -o -name "*.jsonc" -o -name "*.yaml" -o -name "*.yml" \
    -o -name "*.toml" -o -name "*.xml" -o -name "*.ini" -o -name "*.cfg" \
    -o -name "*.conf" -o -name "*.env.example" \
    -o -name "*.md" -o -name "*.mdx" -o -name "*.rst" -o -name "*.txt" \
    -o -name "*.csv" -o -name "*.tsv" -o -name "*.pdf" -o -name "*.docx" \
    -o -name "*.doc" -o -name "*.odt" -o -name "*.pptx" -o -name "*.xlsx" \
  \)

Read the file list. Report the count: "Found {N} eligible files. Uploading..."

---

Step 5 — Upload Each File

For each file, compute REL as its path relative to $FOLDER (strip the leading $FOLDER/ prefix). Upload:

curl -s --request POST \
  --url "https://app.customgpt.ai/api/v1/projects/${AGENT_ID}/sources" \
  --header "Authorization: Bearer ${API_KEY}" \
  --form "file=@${ABSOLUTE_PATH};filename=${REL}"

HTTP 200 or 201 = success. Report progress as you go: ✓ {REL} or ✗ {REL} (HTTP {status}).

Collect totals: $UPLOADED (success) and $FAILED (non-200/201).

---

Step 6 — Save Meta File

cat > "${FOLDER}/.customgpt-meta.json" << EOF
{
  "agent_id": ${AGENT_ID},
  "agent_name": "$(basename $FOLDER)",
  "indexed_folder": "${FOLDER}",
  "created_at": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
EOF

---

Step 7 — Report

Agent created successfully.

• Agent: {agent_name} (ID: {agent_id})
• Folder: {folder}
• Uploaded: {UPLOADED} / {total} files{FAILED > 0 ? ", Failed: {FAILED}" : ""}

CustomGPT.ai is now indexing your files. Run /check-status in a moment to see indexing progress, then /query-agent to start searching.

If any uploads failed, add:

"Some files failed to upload. Run /index-files to retry specific files, or /refresh-agent to re-sync everything."