No bidirectional override characters
Bidirectional embedding, override and isolate controls (U+202A to U+202E, U+2066 to U+2069) make text display in a different order than it is read, so a reviewer sees one thing and the model another.
Check sec-bidi-controls · Poisoning and hidden text · default severity: blocking · rubric 2.1.0
What fails
Bidirectional embedding, override and isolate controls (U+202A to U+202E, U+2066 to U+2069) make text display in a different order than a model reads it, so a reviewer sees one thing and the model another.
The rule
Encoded or obfuscated instructions do not exempt a plugin from these requirements.
How to fix it
Remove the characters. "Copy fixed tool list" strips them.
- Strip the characters from every string in the tool list.
- Keep right-to-left text (Arabic, Hebrew) without explicit overrides: plain text renders correctly.
- Add a CI test that fails on them.
A CI test that fails when the pattern comes back, run against the saved tools/list answer:
TypeScript (node:test)
// tools-list.test.mjs: run `node --test` after saving the answer with
// npx @modelcontextprotocol/inspector --cli <server> --method tools/list > tools.json
import assert from 'node:assert/strict';
import { readFileSync } from 'node:fs';
import { test } from 'node:test';
const answer = JSON.parse(readFileSync('tools.json', 'utf8'));
const tools = answer.tools ?? answer.result?.tools ?? answer;
// Every string and every key in a tool, with its path.
function* strings(value, path) {
if (typeof value === 'string') yield [path, value];
else if (value && typeof value === 'object')
for (const [k, v] of Object.entries(value)) {
yield [`${path}.${k} (key)`, k];
yield* strings(v, `${path}.${k}`);
}
}
const PATTERN = /[\u202A-\u202E\u2066-\u2069]/u;
test('no bidirectional override characters', () => {
const hits = [];
for (const tool of tools) for (const [path, text] of strings(tool, tool.name)) if (PATTERN.test(text)) hits.push(path);
assert.deepEqual(hits, [], 'bidirectional override characters in tool metadata');
});Python (pytest)
# test_tools_list.py: run `pytest` after saving the answer with
# npx @modelcontextprotocol/inspector --cli <server> --method tools/list > tools.json
import json
import re
with open("tools.json", encoding="utf-8") as f:
answer = json.load(f)
tools = answer.get("tools") or answer.get("result", {}).get("tools") if isinstance(answer, dict) else answer
def strings(value, path):
"""Every string and every key in a tool, with its path."""
if isinstance(value, str):
yield path, value
elif isinstance(value, dict):
for k, v in value.items():
yield f"{path}.{k} (key)", k
yield from strings(v, f"{path}.{k}")
elif isinstance(value, list):
for i, v in enumerate(value):
yield from strings(v, f"{path}.{i}")
PATTERN = re.compile(r"[\u202A-\u202E\u2066-\u2069]")
def test_no_bidirectional_override_characters():
hits = [path for tool in tools for path, text in strings(tool, tool["name"]) if PATTERN.search(text)]
assert hits == [], "bidirectional override characters in tool metadata"Questions
Is this the "Trojan Source" problem?
The same mechanism: bidi controls reorder what is displayed. In code it hides logic from reviewers; in tool metadata it hides text from them.
Does right-to-left text need these characters?
Normally no. The Unicode bidi algorithm orders plain Arabic or Hebrew text correctly without explicit controls.
Related checks
- No invisible Unicode tag or variation-selector characters
- No zero-width or other invisible characters
- No control characters in metadata
- All checks in Tool poisoning and hidden text
How this check is graded
It counts toward: Poisoning and hidden text, Anthropic connector directory, OpenAI plugin directory. See the rubric.
Last verified against OpenAI plugin guidelines, section 3.3 on 2026-11-18. The rule above is quoted in the source’s own words. This page as Markdown.
Check your tool list: paste your tools/list, nothing is sent anywhere