skills/synkraai/aiox-core/aiox-devops/Gen Agent Trust Hub

aiox-devops

Warn

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEPERSISTENCEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: Several commands, including *pro-access-grant, *pro-reset-password, and *pro-validate-login, accept password or new_password as plaintext arguments. This practice exposes sensitive user credentials to the agent's execution context, conversation history, and potentially to system logs.
  • [PERSISTENCE]: The skill automatically manages a git pre-push hook at .git/hooks/pre-push. This hook intercepts standard git operations and enforces agent-specific restrictions based on environment variables, representing a persistent modification of the user's development environment.
  • [DYNAMIC_EXECUTION]: The agent is configured to execute local JavaScript scripts (asset-inventory.js, path-analyzer.js, migrate-agent.js) and a unified-activation-pipeline.js fallback script during activation. Additionally, it dynamically constructs and executes shell commands for the CodeRabbit CLI, including platform-specific wrappers for Windows (WSL).
  • [INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface for indirect prompt injection by consuming untrusted data from the repository, such as git commit messages, package.json metadata, and handoff artifacts in .aiox/handoffs/. Malicious content in these files could influence the agent's automated workflows, such as pull request creation or release generation.
  • Ingestion points: Git status output, repository metadata files, and YAML handoff artifacts in .aiox/handoffs/.
  • Boundary markers: Absent; the agent interprets this data directly to determine suggested next steps and workflow chains.
  • Capability inventory: Includes git push, gh pr create, npm run scripts, and execution of local .js files via Node.js.
  • Sanitization: No evidence of sanitization or filtering for the external data ingested from the repository.
  • [COMMAND_EXECUTION]: The skill executes various potentially dangerous system commands including npm run build/test/lint, git push --force, and GitHub CLI operations. While these are intended for DevOps tasks, they operate with the agent's authority over the remote repository.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 11, 2026, 07:56 PM