epic-conductor

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell scripts and tools defined within the repository's configuration profile, such as scripts/agent-identity.sh for identity management and scripts/gh-upload-media.sh for asset handling.
  • [REMOTE_CODE_EXECUTION]: The skill orchestrates a fleet of specialized agents by invoking sub-skills such as agentloop:design-review and agentloop:pr-review, delegating implementation and verification tasks to isolated worker contexts.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface by processing untrusted content from GitHub issues and pull request comments to drive its workflow.\n
  • Ingestion points: GitHub issue content and comments (gh issue view) are used to brief worker agents.\n
  • Boundary markers: The skill mandates isolation: \"worktree\" for worker agents to maintain environment separation.\n
  • Capability inventory: The orchestration logic has access to Bash, Write, Edit, Agent, and Skill tools.\n
  • Sanitization: No explicit sanitization of external issue content is specified, with the system relying on isolated contexts and model reasoning.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 10:26 PM
Security Audit — agent-trust-hub — epic-conductor