No encoded text blocks in metadata
A long base64 or hex block in a description is unreadable to a reviewer but not to a model, so it can hide instructions; if it decodes to readable text the checker blocks it, and both platforms reject encoded instructions.
Check sec-encoded-blob · Poisoning and hidden text · default severity: attention (or blocking, depending on the case) · rubric 2.1.0
What fails
A long base64 or hex run in a description is unreadable to a reviewer but not to a model. If it decodes to readable text it is blocking, and the message shows the decoded start.
The rule
Encoded or obfuscated instructions do not exempt a plugin from these requirements.
How to fix it
Remove the encoded block. Put examples in plain text.
- Remove encoded blocks from descriptions and titles.
- Put examples in plain text; link to binary samples instead of inlining them.
- Add a CI test for long base64 runs.
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 = /[A-Za-z0-9+/]{60,}={0,2}/;
test('no long encoded blocks', () => {
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, [], 'encoded text hides content from reviewers');
});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"[A-Za-z0-9+/]{60,}={0,2}")
def test_no_long_encoded_blocks():
hits = [path for tool in tools for path, text in strings(tool, tool["name"]) if PATTERN.search(text)]
assert hits == [], "encoded text hides content from reviewers"Questions
What about a hash or an ID in a description?
A 64-character hex hash will match the simple CI pattern above; the checker itself only blocks blocks that decode to readable text. Move IDs out of descriptions anyway.
Can a model really decode base64 in a description?
Yes. Current models read base64 fluently, which is why OpenAI says encoded instructions do not exempt a plugin from its rules.
Related checks
- No invisible Unicode tag or variation-selector characters
- No HTML comments in metadata
- No instructions fetched from outside sources
- 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