
from devpilot-agents31
CLI wrapper for multi-backend AI code tasks—run, orchestrate, and parallelise jobs across Codex, Claude, Gemini and OpenCode backends with file references and s
CodeAgent provides a CLI wrapper for executing code-focused tasks across multiple AI backends (Codex, Claude, Gemini, OpenCode). It standardises invocation patterns, supports @file references, parallel execution with per-task backend selection, and produces structured summaries or full outputs for debugging. The skill documents invocation syntaxes, backend recommendations, concurrency controls, and security flags.
Use CodeAgent when automating complex code analysis, large-scale refactors, multi-file changes, or orchestration that benefits from backend specialization (e.g., Codex for implementation & review, Gemini for UI components). It's useful in CI automation, developer tooling, and scripted agent-driven workflows where per-task backend tuning improves results.
Agents and automation runners that can execute shell CLIs and manage tmux sessions, plus LLM backends (Codex, Claude, Gemini) when integrated. Ideal for developer-oriented agents and orchestration pipelines.
CodeAgent Wrapper is a CLI orchestration skill for dispatching AI code tasks across Codex, Claude, Gemini, and OpenCode backends with parallel execution and tmux visualization. No bundled scripts to test — purely a SKILL.md instruction set. Security concerns include documented --dangerously-skip-permissions and CODEX_BYPASS_SANDBOX flags that disable safety checks, though both are opt-in. Well-documented with clear examples but monolithic structure with no script or reference separation.
No scripts present for runtime testing. Security deductions for documenting patterns that disable safety mechanisms (sandbox bypass, permission skipping), though these are presented as opt-in controls rather than defaults. The skill is essentially a well-written CLI usage guide with no executable code to audit.