Answer: GitHub mitigates prompt injection risks by filtering hidden characters before passing user input to Copilot cloud agent, so text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent. Status: Confirmed Entry: tool/github-copilot/permissions-and-security/mitigate-prompt-injection Revision: 2 Digest: sha256:fd17199c8c8573e0 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:21:35Z Last changed: 2026-11-18T12:51:56Z (example: — → text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent) Last verified: 2026-11-20T12:58:10Z Next check: 2026-11-21T12:58:10Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of mitigation | GitHub filters hidden characters before passing user input to Copilot cloud agent | [1] | 2026-11-20 example | text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent | [1] | 2026-11-20 Sources: [1] GitHub. "Risks and mitigations for GitHub Copilot cloud agent - GitHub Docs". https://docs.github.com/en/copilot/concepts/security-governance-and-network-settings/risks-and-mitigations Role: primary. Accessed: 2026-11-20T12:58:10Z Quote: "To mitigate this risk, GitHub filters hidden characters before passing user input to Copilot cloud agent" (Supports: mitigation) Quote: "For example, text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent." (Supports: example) Snapshot SHA-256: 26bd8b94b256b37b1cc460150dd7125edd8d0393433bcb5f7a341c9ae9df6e7e (captured 2026-11-18T12:51:55Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. On this check the text of [1] was unchanged since 2026-11-18T12:51:56Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T12:58:10Z. No independent second source is configured for this entry Last successful check: 2026-11-20T12:58:10Z Last attempt: 2026-11-20T12:58:10Z Notes and limitations: Extracted automatically from a single official source (section "AI prompts can be vulnerable to injection"). Related: https://latestn.com/tool/github-copilot/permissions-and-security/network-access-permissions Who can manage GitHub Copilot network access for an organization? https://latestn.com/tool/github-copilot/permissions-and-security/org-vs-individual-space-sharing What is the difference between sharing organization-owned and individual-owned GitHub Copilot Spaces? https://latestn.com/tool/github-copilot/permissions-and-security/pass-token-to-sdk How do I pass a GitHub OAuth token to the GitHub Copilot SDK? https://latestn.com/tool/github-copilot/permissions-and-security/mcp-registry-who-can-use Who can use the GitHub Copilot MCP registry restriction feature? https://latestn.com/tool/github-copilot/permissions-and-security/mcp-security-considerations What security considerations should I keep in mind when using MCP with GitHub Copilot agent mode? https://latestn.com/tool/github-copilot/permissions-and-security/mcp-server-usage-company How does GitHub Copilot control MCP server usage in a company? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-18T12:51:56Z | changed | example: — → text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent | [1] 2026-11-18T12:51:56Z | changed | mitigation: — → GitHub filters hidden characters before passing user input to Copilot cloud agent | [1] 2026-11-18T12:51:56Z | changed | Example of filtered input: text entered as an HTML comment in an issue or pull request comment is not passed to Copilot cloud agent → — | [1] 2026-11-18T12:51:56Z | changed | Mitigation: GitHub filters hidden characters before passing user input to Copilot cloud agent → — | [1] 2026-11-13T03:21:35Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/github-copilot/permissions-and-security/mitigate-prompt-injection/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/github-copilot/permissions-and-security/mitigate-prompt-injection POST: https://latestn.com/dispute/tool/github-copilot/permissions-and-security/mitigate-prompt-injection with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/github-copilot/permissions-and-security/mitigate-prompt-injection?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How can I mitigate prompt injection risks when using GitHub Copilot cloud agent? Revision 2, last verified 2026-11-20T12:58:10Z. https://latestn.com/tool/github-copilot/permissions-and-security/mitigate-prompt-injection (accessed 2026-11-21T09:57:29Z). @misc{LatestN_tool_github_copilot_permissions_and_security_mitigate_prompt_injection_r2, title = {How can I mitigate prompt injection risks when using GitHub Copilot cloud agent?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/github-copilot/permissions-and-security/mitigate-prompt-injection}}, note = {Revision 2; digest sha256:fd17199c8c8573e0; last verified 2026-11-20T12:58:10Z; accessed 2026-11-21T09:57:29Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/github-copilot/permissions-and-security/mitigate-prompt-injection