1k-create-pr

Warn

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands (git and gh) using variables such as <branch-name>, <type>, and <description> that are dynamically generated from conversation context. Without proper escaping, this could lead to command injection if malicious strings are introduced into the conversation history.\n- [DATA_EXFILTRATION]: The workflow utilizes git add . to stage changes before pushing to a remote server. This behavior risks the accidental commitment and exfiltration of sensitive local files, such as environment variables or private keys, if they are not correctly excluded by a .gitignore file.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it explicitly instructs the agent to populate PR metadata based on the full conversation history without providing sanitization or boundary markers.\n
  • Ingestion points: Full conversation history used to extract Intent, Root Cause, and Design Decisions.\n
  • Boundary markers: Absent; no delimiters are used to separate untrusted conversation content from the system instructions.\n
  • Capability inventory: The skill can perform repository modifications (git commit), code uploads (git push), and automated merges (gh pr merge).\n
  • Sanitization: None; the skill lacks any mechanism to validate or filter the extracted context before it is inserted into command arguments or the PR body.\n- [COMMAND_EXECUTION]: The use of gh pr merge --auto --squash enables automated merging of changes. When combined with AI-generated PR content and automated pushing, this increases the risk of malicious code being merged into the primary branch without a manual security review if the agent is successfully manipulated.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 4, 2026, 04:18 AM