agent-orchestrator

Pass

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs agents to perform standard repository-level operations such as committing changes with git and executing validation commands (tests, lint, build, typecheck) discovered within the project's configuration files (e.g., package.json, pyproject.toml).
  • [PROMPT_INJECTION]: The orchestrator is exposed to indirect prompt injection risks as it and its sub-agents ingest and process data from external repositories.
  • Ingestion points: The agent reads README files, source code, and configuration metadata during audit and implementation phases (SKILL.md steps 5, 7, and 9).
  • Boundary markers: Not explicitly defined; the skill does not require the use of delimiters or 'ignore instructions' warnings when processing untrusted content.
  • Capability inventory: Sub-agents like 'implementer' have broad capabilities including writing to the filesystem and executing shell commands.
  • Sanitization: Absent; the skill does not specify validation or filtering for external repository content.
  • [EXTERNAL_DOWNLOADS]: The workflow mentions the use of tools for retrieving external content from URLs (e.g., $web-fetch-md) to support research and documentation tasks within the engineering process.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 26, 2026, 12:13 AM