coder-loop

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes platform-native commands (codex plugin) to manage its lifecycle and coordinates specialized subagents using the fork_turns mechanism. These operations use standard host features as intended for multi-agent orchestration.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the discovery and installation of optional components from the author's verified marketplace (transcendr/slopware-skills). These operations are transparent, target vendor-owned resources, and require explicit user consent before execution.
  • [PROMPT_INJECTION]: As a development orchestration tool, the skill processes external data such as source code, configuration files, and worker artifacts which could contain instructions designed to influence agent behavior.
  • Ingestion points: Repository source files, configuration files, and reports generated by implementation and review subagents.
  • Boundary markers: Employs strict role contracts and bounded "candidate boundaries" to isolate the context of subagent tasks.
  • Capability inventory: Spawns subagents with variable reasoning efforts, performs file system operations, and manages platform-level plugin installations.
  • Sanitization: Relies on structural process isolation and mandatory independent review to mitigate risks from malicious content; no explicit programmatic sanitization of ingested code is specified.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 07:34 AM
Security Audit — agent-trust-hub — coder-loop