miro-code-review
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFEDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions require the agent to build and run shell commands using variables such as $PR_NUMBER, $OWNER, and $REPO derived from user input or external URLs. The lack of explicit sanitization for these variables poses a command injection risk if the input contains malicious shell metacharacters.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from Pull/Merge Requests, including titles, descriptions, and code diffs. It lacks boundary markers or instructions to disregard embedded commands in this data, which could allow an attacker to influence the agent's analysis or subsequent actions through malicious PR content.
- [CREDENTIALS_UNSAFE]: The workflow involves using sensitive environment variables like $GITHUB_TOKEN and $GITLAB_TOKEN, as well as the ~/.netrc file, for authentication. While functional for the skill's purpose, this highlights a sensitive data access capability that must be handled securely.
- [DATA_EXFILTRATION]: The skill fetches source code and metadata from repositories and transmits analysis results to Miro. This data flow from the code forge to the Miro platform is the primary function of the skill but involves handling sensitive source information.
Audit Metadata