[Yashwantgokul]

Summary

Adds five typo-based prompt-injection payload examples under the existing Instruction Obfuscation category, covering character transposition, omission, duplication, spacing variation, and phonetic obfuscation.

Motivation / Security-Research Context

Some prompt-injection detection systems may rely on exact keyword matching or pattern-based filters. Intentional spelling and spacing variations can help evaluate whether these systems recognize instruction-like content despite superficial textual changes. These examples are intended for defensive evaluation and authorized security testing.

Affected Category

Prompt-Injection → Instruction Obfuscation in prompt-injection-payloads.txt.

Duplicate Check

• I checked the target category (and related ones) for existing/similar entries.
• Notes: Reviewed the target and related categories for overlap with existing leetspeak, Unicode, case-variation, and other text-obfuscation payloads.

Validation Performed

Manually reviewed the proposed examples for relevance to intentional misspelling and typo-based obfuscation. Confirmed that each example illustrates a distinct variation. No live-system testing is claimed.

Source Attribution and References

N/A — original examples drafted for this contribution.

AI-Assisted Contribution Disclosure

• No AI assistance used.

• AI tools assisted with this contribution.

• Tool/model: ChatGPT (GPT-6)

• Provider: OpenAI

• What it helped with: Brainstorming typo-based payload examples and drafting the PR description.

• Human validation performed: Reviewed the proposed examples and their technique labels; checked for duplicates and verified formatting.

Licensing and Originality

• This contribution is my own work, or is properly attributed to its original source.
• I have the right to submit this content under the project's MIT License.
• This content does not include credentials, secrets, or private/customer data.

Responsible Use Confirmation

• I have read [DISCLAIMER.md](../DISCLAIMER.md) and this contribution is consistent with authorized-testing-only use.

[github-actions]

✅ No duplicate payload entries detected

Checked 5 normalized entries from added/modified payload files on the latest push.

Posted automatically by the Duplicate Payload Check workflow.

[0x5t4l1n]

Thanks for the contribution! The five typo-based payloads are relevant and non-duplicate. Please remove the Markdown bullet prefixes (- ) from the new payload entries to maintain consistency with the existing TXT payload format. Once updated, this PR should be ready to merge.

[Yashwantgokul]

changes made

[0x5t4l1n]

Thanks for the fixes! Changes look good, and the duplicate check passed. LGTM, approved for merge!