Delegate bounded coding tasks to external CLI agents with structured verification and evidence collection.
npx skills add https://github.com/logicrw/agpair --skill agpair
Please help me install this Agent Skill. Skill: agpair Source: https://github.com/logicrw/agpair/tree/main/skills/Claude Command: npx skills add https://github.com/logicrw/agpair --skill agpair
The agpair Skill unit addresses the challenge of managing complex, non-trivial coding tasks by leveraging external CLI agents for delegation, verification, and evidence collection, enhancing Claude Code's capabilities.
To start a task with agpair, use the following command:
agpair task start --repo-path "$REPO" --executor grok-cli --task-kind quick_review --wait-policy lease --authorization-profile local_readonly --completion-policy report --body "$BRIEF"
High-intent search queries and answers about installing and using this skill.
Task delegation to external CLI agents from Claude Code is managed by sending bounded work to executors like grok-cli. The agpair Skill handles the lifecycle, sending tasks out and collecting structured receipts for execution verification.
Verifying code review results from CLI agents involves validating structured receipts and supporting state-aware retries. The system checks the evidence returned by executors to ensure the delegated work meets the required completion policies.
Yes, you can use grok-cli and antigravity-cli as external executors for task delegation. The Skill supports multiple CLI agents, allowing you to configure different executors based on the specific task kind and required authorization profile.
Evidence collection for delegated code tasks gathers implementation candidates, alternative analysis, and verification evidence. This structured evidence layer captures the output and state of external CLI agent executions for review.
Yes, Claude Code integration is required to run CLI agent task delegation. The system acts as an evidence and lifecycle manager that orchestrates external CLI executors specifically within the Claude Code environment.
Limitations of using CLI agents for code verification include the dependency on external executor availability and the need to define bounded work. Tasks require explicit authorization profiles and completion policies to ensure safe execution and proper fallback handling.
Khang5687Route implementation tasks to executors and Claude subagents based on complexity.
xz1220Run a persistent iteration loop with tests, build, lint, and architect validation until completion.
Happyverse-TeamManage runtime tasks and memories via Ralph CLI during orchestration runs.
logicrwDelegate coding tasks to external CLI agents with Codex as controller.