Detection queries Not classified

Radware hidden prompt injection targeting GitHub maintainer bots

Radware researchers Zvika Babo and Gabi Nakibly (Nov 5, 2026) document hidden prompt injection techniques targeting GitHub maintainer bots processing issues and pull requests. Attackers hide instructions in…

What happened

Radware researchers Zvika Babo and Gabi Nakibly (Nov 5, 2026) document hidden prompt injection techniques targeting GitHub maintainer bots processing issues and pull requests. Attackers hide instructions in HTML comments, invisible Unicode characters, and CSS display tricks. Examples documented: (1) google-gemini/gemini-cli#28575 - deduplication agent told to use run_shell_command to write base64-encoded GITHUB_TOKEN into GITHUB_ENV; (2) openai/codex#16733 - labeler told to curl webhook.site and dump environment variables; (3) microsoft/agent-framework#7177 - fake triage gate with allow_triage=true, gate_result=pass markers; (4) blondesean/Kings_Herald#21 - hidden backdoor instruction in feature request. Attackers study bot tool names and output field formats to forge steering markers. Hidden instructions can bypass human review while manipulating bot behavior. Radware provides examples repository at github.com/rdwr-gabin/github-ipi-in-the-wild. No specific threat actor attribution; confirmed technique active in wild.

GitHub triage, review, and CI bots manipulated via hidden HTML comments, invisible Unicode, and CSS tricks in issues/PRs to execute unauthorized actions

Read the original sourceAdditional source: Radware: GitHub IPI in the Wild - Examples Repository

How it happened

Select a stage to explore its defenses
Detect this stage
Detect this stage
Detect this stage
Select a stage · Drag to explore · + / − to zoom · Dashed edges are loopsSelect a stage · Arrows show reported connectionsReported sequence
Read the flow as text & sources

A catalog reconstruction of the supplied source account. Unknown steps remain explicit; an edge describes the reported relationship between steps.

  1. 01 · Study bot tools and output formats

    Attacker analyzes target repository bot behavior, tool names, and expected output field formats

    Source for step 1
  2. 02 · Craft hidden prompt injection

    Embed instructions in HTML comments, invisible Unicode, or CSS-hidden elements

    Hunt: An agent ingests external content or an untrusted session identifier before a tool action or configuration change outside the approved task.

    Requires application instrumentation and a custom field mapping. Input exposure does not prove that the model followed instructions, executed a tool or bypassed approval. Avoid collecting sensitive prompt bodies unless authorized.

    Source for step 2
  3. 03 · Submit malicious issue or PR

    Attacker creates GitHub issue or PR containing hidden instructions

    Hunt: A token or agent reads private material and creates a public issue, repository, commit or visibility change inconsistent with the approved task.

    Git-provider audit completeness depends on product and plan. Not every file read or issue body is present in organization audit logs.

    Source for step 3
  4. 04 · Bot processes issue/PR

    Triage, review, or CI bot reads issue/PR including hidden injection

    Hunt: Untrusted content or a changed tool definition precedes a file read, command, or remote call outside the user's intended task.

    Process logs cannot recover model intent. Remote connectors and in-process interpreters may produce no local child process. Prompt content alone is not a reliable signature.

    Source for step 4
  5. 05 · Bot executes injected instructions

    Bot writes GITHUB_TOKEN to env, curls webhook, or accepts fake triage gate

    Source for step 5

Why it matters

Reported entry point: Malicious GitHub issues or pull requests with hidden prompt injection. Reported impact: Bot writes GITHUB_TOKEN to env, curls webhook, or accepts fake triage gate; reported targets include GitHub repositories using AI-powered triage, review, and CI bots; examples include Google Gemini, OpenAI, Microsoft repositories; applicability to your setup has not been assessed, so verify affected versions, permissions, and configuration in the original report.

Read the original report

What you can do

Protection & prevention

Strip HTML comments before bot processing; sanitize invisible Unicode from issues/PRs; validate bot tool output fields against expected schema; implement bot action approval gates; monitor for suspicious run_shell_command, curl webhook, environment variable access; restrict bot permissions to minimum required; log all bot actions for audit; educate maintainers on prompt injection risks; implement bot behavioral analysis

Catalog defensive guidance based on the reported behavior. Confirm scope and current mitigations in the original source.

Review the original guidance Additional source 2
Investigate each phase

Detection queries

Draft query starters grouped by the attack stages where they apply.

Download stage pack

Start with the logs you collect. Each hunt identifies its required visibility, field mapping and limits. Queries are catalog-authored drafts using reusable behavior profiles; test with known activity and your normal baseline before alerting.

02Craft hidden prompt injectionUntrusted content enters an agent sessionCustom schema · Draft

Embed instructions in HTML comments, invisible Unicode, or CSS-hidden elements

What to look for

An agent ingests external content or an untrusted session identifier before a tool action or configuration change outside the approved task.

Visibility: Application / agent audit records

Required logs

  • Input provenance, source type and request/session identifier
  • Captured tool responses or retrieved content where collection is approved
  • Subsequent tool calls, configuration changes and approval decisions

Coverage limit. Requires application instrumentation and a custom field mapping. Input exposure does not prove that the model followed instructions, executed a tool or bypassed approval. Avoid collecting sensitive prompt bodies unless authorized.

Custom-schema candidate for input exposure only. Map agent.input.ingest and agent.session.resume to your application request and session logs. CorrelationId must link subsequent actions. EventType values are proposed normalization names, not native vendor events.

Sigma
Draft SigmaDownload Sigma
title: Hunt - Untrusted content enters an agent session
id: 2d89d7ae-9d44-4dc6-a95b-f9c8de7b8352
status: experimental
description: An agent ingests external content or an untrusted session
  identifier before a tool action or configuration change outside the approved
  task.
author: MagicSword OpenAttacks catalog
date: 2026-11-04
falsepositives:
  - Authorized activity matching this broad hunt; baseline and investigate
    before alerting.
level: low
logsource:
  category: application
  product: openattacks_audit
  definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
    Target and CorrelationId from the required source logs first.
detection:
  selection:
    EventType:
      - agent.input.ingest
      - agent.session.resume
  condition: selection
KQL
Draft KQLDownload KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("agent.input.ingest", "agent.session.resume")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated desc
SPL
Draft SPLDownload SPL
sourcetype="openattacks:audit"
EventType IN ("agent.input.ingest", "agent.session.resume")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _time
Investigation steps
  1. Join the ingested input to later actions by session or request ID; preserve the original content as evidence.
  2. Compare the resulting action with the approved task and baseline. Receiving suspicious text alone is not successful injection.
03Submit malicious issue or PRPrivate access → public repository artifactCustom schema · Draft

Attacker creates GitHub issue or PR containing hidden instructions

What to look for

A token or agent reads private material and creates a public issue, repository, commit or visibility change inconsistent with the approved task.

Visibility: Defender-side logs where collected

Required logs

  • Git-provider organization / enterprise audit
  • Repository and token permission inventory
  • Agent tool and secret-scanning / DLP records

Coverage limit. Git-provider audit completeness depends on product and plan. Not every file read or issue body is present in organization audit logs.

Custom-schema candidates, not out-of-box detectors. Map your application / cloud / platform logs to EventType, Actor, Asset, Action, Target and CorrelationId. KQL uses a custom OpenAttacksAudit_CL table; SPL uses openattacks:audit. EventType values are proposed normalization names, not native vendor event names. Correlation and environment-specific alert logic remain to be built.

Sigma
Draft SigmaDownload Sigma
title: Hunt - Private access → public repository artifact
id: b7b2d3f0-8170-4225-accb-c20633364ef4
status: experimental
description: A token or agent reads private material and creates a public issue,
  repository, commit or visibility change inconsistent with the approved task.
author: MagicSword OpenAttacks catalog
date: 2026-10-23
falsepositives:
  - Authorized activity matching this broad hunt; baseline and investigate
    before alerting.
level: low
logsource:
  category: application
  product: openattacks_audit
  definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
    Target and CorrelationId from the required source logs first.
detection:
  selection:
    EventType:
      - repository.public.write
      - repository.visibility.change
  condition: selection
KQL
Draft KQLDownload KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("repository.public.write", "repository.visibility.change")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated desc
SPL
Draft SPLDownload SPL
sourcetype="openattacks:audit"
EventType IN ("repository.public.write", "repository.visibility.change")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _time
Investigation steps
  1. Join access and public-write actions by actor / token and time; inspect visibility changes and newly created repositories.
  2. Verify whether the artifact contains sensitive data using approved scanning, retaining original source qualifications.
04Bot processes issue/PRUnexpected agent tool useCustom schema · Draft

Triage, review, or CI bot reads issue/PR including hidden injection

What to look for

Untrusted content or a changed tool definition precedes a file read, command, or remote call outside the user's intended task.

Visibility: Defender-side logs where collected

Required logs

  • Agent tool-call audit records with session / tool / destination
  • Tool-definition and approval changes
  • Endpoint process, file-access and network evidence where available

Coverage limit. Process logs cannot recover model intent. Remote connectors and in-process interpreters may produce no local child process. Prompt content alone is not a reliable signature.

Custom-schema candidates, not out-of-box detectors. Map your application / cloud / platform logs to EventType, Actor, Asset, Action, Target and CorrelationId. KQL uses a custom OpenAttacksAudit_CL table; SPL uses openattacks:audit. EventType values are proposed normalization names, not native vendor event names. Correlation and environment-specific alert logic remain to be built.

Sigma
Draft SigmaDownload Sigma
title: Hunt - Unexpected agent tool use
id: c709e384-0e11-479e-ab4d-3a06e2431e4b
status: experimental
description: Untrusted content or a changed tool definition precedes a file
  read, command, or remote call outside the user's intended task.
author: MagicSword OpenAttacks catalog
date: 2026-10-23
falsepositives:
  - Authorized activity matching this broad hunt; baseline and investigate
    before alerting.
level: low
logsource:
  category: application
  product: openattacks_audit
  definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
    Target and CorrelationId from the required source logs first.
detection:
  selection:
    EventType:
      - agent.tool.call
      - agent.tool.definition.change
  condition: selection
KQL
Draft KQLDownload KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("agent.tool.call", "agent.tool.definition.change")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated desc
SPL
Draft SPLDownload SPL
sourcetype="openattacks:audit"
EventType IN ("agent.tool.call", "agent.tool.definition.change")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _time
Investigation steps
  1. Join input provenance, tool descriptions and calls by session ID and compare the action to the approved task.
  2. Inspect child processes and destinations; record whether execution was in-process, local, or through a remote SaaS connector.
Technical detailsTooling, impact & MITRE ATLAS

Attack context

AI's role
Agent / MCP as target
Initial access
Malicious GitHub issues or pull requests with hidden prompt injection

Tooling as reported

AI products, models and artifacts

  • GitHub triage bots
  • GitHub review bots
  • GitHub CI/CD automation bots
  • Repository automation agents

Other tooling

  • HTML comments
  • Invisible Unicode characters
  • CSS display manipulation
  • webhook.site exfiltration

Targets and reported impact

Targets / sectors
GitHub repositories using AI-powered triage, review, and CI bots; examples include Google Gemini, OpenAI, Microsoft repositories
Reported scale / exposure
Multiple confirmed examples across major repositories (Google, OpenAI, Microsoft); active Nov 2026
Target geography
Multiple regions

Reported attribution

Unattributed technique; confirmed active in wild with documented examples

Structured attribution

Actor
Unattributed technique
Country
Unknown
Motivation
unknown
Attributed by
Radware
Attribution confidence
low

Detection rules

  • Execution via GitHub Actions RunnerELASTICView rule

    Detects malicious code execution via GitHub Actions bot manipulation

  • Tampering with RUNNER_TRACKING_ID in GitHub Actions RunnersELASTICView rule

    Detects GITHUB_ENV and GITHUB_TOKEN manipulation via bot prompt injection

  • GitHub bot security controlsGUIDANCE

    Strip HTML comments before bot processing; sanitize invisible Unicode (U+200B, U+200C, U+200D, U+FEFF); detect CSS display:none in issues/PRs; monitor run_shell_command usage; alert on curl webhook.site from bots; restrict GITHUB_TOKEN write access

Activity timeline

First seen
2026-10-01
Last updated
2026-11-05

Updates

  • 2026-11-05 — Radware

    Zvika Babo and Gabi Nakibly publish analysis of hidden prompt injection targeting GitHub maintainer bots; examples repository documents real-world cases from Google, OpenAI, Microsoft repositories

Source defensive notes

Strip HTML comments before bot processing; sanitize invisible Unicode from issues/PRs; validate bot tool output fields against expected schema; implement bot action approval gates; monitor for suspicious run_shell_command, curl webhook, environment variable access; restrict bot permissions to minimum required; log all bot actions for audit; educate maintainers on prompt injection risks; implement bot behavioral analysis

MITRE ATLAS

AI-specific catalog annotations based on the reported behavior; these are not MITRE case-study assignments.

AML.T0051LLM Prompt Injection

Hidden HTML comments, invisible Unicode, and CSS tricks inject instructions into GitHub bot processing Source

AML.T0053AI Agent Tool Invocation

Bots manipulated to execute run_shell_command, curl webhooks, write GITHUB_TOKEN to environment Source

AML.T0055Unsecured Credentials

Attackers study bot tool names/output fields to forge valid steering markers (allow_triage, gate_result) Source

Indicators of compromise

domains

  • webhook.site

ips

    hashes

      urls

      • https://github.com/google-gemini/gemini-cli/issues/28575
      • https://github.com/openai/codex/issues/16733
      • https://github.com/microsoft/agent-framework/issues/7177
      • https://github.com/blondesean/Kings_Herald/issues/21
      • https://github.com/rdwr-gabin/github-ipi-in-the-wild

      Sources & research credit

      2 sources

      Research by Zvika Babo, Gabi Nakibly · Radware

      1. Radware: Prompt Injecting Coding Agents www.radware.com
      2. Radware: GitHub IPI in the Wild - Examples Repository github.com

      Radware blog (Nov 5, 2026) primary source; examples repository github.com/rdwr-gabin/github-ipi-in-the-wild provides proof-of-concept demonstrations; links to hackerbot-claw-github-actions-2026-03 and forever-agents-md-hijack catalog entries for related GitHub AI agent attacks

      Record & downloads
      Source record name
      Radware hidden prompt injection targeting GitHub maintainer bots
      Entry ID
      radware-github-prompt-injection-bots-2026-11
      Catalog review
      November 10, 2026
      Category
      attacks-on-ai
      AI involvement confidence
      confirmed