Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Studio: accept a SKILL.md saved with a UTF-8 byte order mark (#12195)

Mohammad Hijjawia41017fe03df
diff --git a/studio/backend/core/inference/skills.py b/studio/backend/core/inference/skills.py
index b8b52aac0fc..02605ccc5c3 100644
--- a/studio/backend/core/inference/skills.py
+++ b/studio/backend/core/inference/skills.py
@@ -285,7 +285,7 @@ def _split_skill_markdown(raw: bytes) -> tuple[dict, str]:
if len(raw) > MAX_SKILL_MD_BYTES:
raise SkillError("SKILL.md exceeds the 512 KB limit.")
try:
- text = raw.decode("utf-8")
+ text = raw.decode("utf-8-sig")
except UnicodeDecodeError as exc:
raise SkillError("SKILL.md must be UTF-8 text.") from exc
lines = text.splitlines(keepends = True)
diff --git a/studio/backend/tests/test_skills.py b/studio/backend/tests/test_skills.py
index 814c38a8943..e3ff92edb97 100644
--- a/studio/backend/tests/test_skills.py
+++ b/studio/backend/tests/test_skills.py
@@ -1,6 +1,7 @@
# SPDX-License-Identifier: AGPL-3.0-only
# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+import codecs
import json
import os
import sys
@@ -892,6 +893,19 @@ def test_indented_separator_inside_a_block_scalar_stays_in_the_frontmatter(isola
assert record["description"] == "Use for reports.\n---\nAlso for summaries."
+def test_skill_manifest_saved_with_a_utf8_byte_order_mark_is_valid(isolated_skills):
+ home, _ = isolated_skills
+ root = _write_skill(home, "agents", "marked", body = "Follow the steps.")
+ manifest = root / "SKILL.md"
+ manifest.write_bytes(codecs.BOM_UTF8 + manifest.read_bytes())
+
+ record = next(item for item in skills.list_skills(home = home) if item["name"] == "marked")
+
+ assert record["valid"] is True
+ assert record["description"] == "Use this skill for testing."
+ assert skills.read_skill_manifest("marked", home = home)["instructions"] == "Follow the steps."
+
+
def test_read_skill_tool_applies_defaults_for_null_arguments(isolated_skills, monkeypatch):
from core.inference import tools as tools_module