gh-issues

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXPOSURE]: The skill accesses sensitive local configuration files at ~/.openclaw/openclaw.json and /data/.clawdbot/openclaw.json to retrieve the GH_TOKEN for GitHub authentication.
  • [COMMAND_EXECUTION]: The instructions use node -e to execute inline JavaScript for parsing JSON data from configuration files at runtime.
  • [DATA_EXFILTRATION]: The skill includes a --notify-channel parameter that allows sending information about pull requests to an external Telegram channel based on a user-provided ID.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it processes untrusted data from GitHub issue titles, bodies, and pull request comments. This content is passed to sub-agents with shell execution and repository write capabilities.
  • Ingestion points: Fetches issue metadata in Phase 2 and review comments/PR bodies in Phase 6.
  • Boundary markers: The sub-agent prompt uses <issue> and <review_comments> XML-style tags to delimit untrusted data.
  • Capability inventory: Sub-agents can execute arbitrary shell commands (git, curl, grep), modify the local filesystem, push code to remote repositories, and open pull requests.
  • Sanitization: No explicit sanitization or filtering is performed on the untrusted data before it is interpolated into the sub-agent prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 01:08 AM
Security Audit — agent-trust-hub — gh-issues