review-changes

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from external sources and uses it to guide the agent's behavior.\n
  • Ingestion points: The skill fetches PR metadata (titles, descriptions, comments) via gh pr view and Jira ticket details via acli jira workitem view in SKILL.md.\n
  • Boundary markers: The instructions lack explicit delimiters or warnings for the agent to ignore instructions contained within the fetched external content.\n
  • Capability inventory: The agent has the ability to post comments back to GitHub using GraphQL mutations and can spawn sub-agents to process data.\n
  • Sanitization: There is no evidence of sanitization or validation of the external content before it is processed or used in subsequent steps.\n- [COMMAND_EXECUTION]: The skill generates shell commands for GitHub and Atlassian CLIs by interpolating variables such as PR numbers, repository names, and Jira IDs.\n
  • Evidence: Commands like gh pr view <number> --repo <owner/repo> and acli jira workitem view <JIRA-ID> are constructed using data derived from user input or PR metadata.\n
  • Risk: If the source metadata contains shell metacharacters and the execution environment does not properly escape these values, it could allow for arbitrary command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 11:04 PM