Executes approved implementation plans task-by-task with enforced review gates and resumable session state.
npx skills add https://github.com/greglas75/zuvo --skill execute-greglas75
Please help me install this Agent Skill. Skill: execute Source: https://github.com/greglas75/zuvo/tree/main/skills/execute Command: npx skills add https://github.com/greglas75/zuvo --skill execute-greglas75
High-intent search queries and answers about installing and using this skill.
Approve a plan document with status Approved, then invoke the execute skill. It runs each task through an implementer, spec reviewer, quality reviewer, adversarial review, and acceptance verification before committing, with no manual intervention between tasks.
It rewrites execution-state.md after every task with completed tasks, retry counts, and the next task pointer. On restart it reads this file and resumes directly at the next task, and an optional cron watchdog auto-resumes after API errors or rate-limit stalls.
Yes, but only when tasks are mutually independent in the dependency graph and touch no production or test files in common. Tasks sharing a file are always serialized to prevent lost edits, and each parallel task still runs the full review cycle.
Yes. On runtimes without agent dispatch, such as Cursor, it falls back to single-agent mode where the same agent performs implementer, spec review, quality review, adversarial, and acceptance passes as sequential checkpoints with explicit gate markers.
The task is not committed. Depending on the plan's declared failure strategy it halts, skips and continues, or runs degraded, and the failure is recorded in telemetry. A task missing any gate marker stays IN_PROGRESS rather than being committed.
thiansitGenerate implementation plans with TDD steps, commands, and saved file paths.
fcastrilloCloses a project epic by finalizing gradebook entries, aggregating statistics, and reopening the next epic.
xriuAudits codebases and writes self-contained implementation plans for other agents to execute.
KbediakoConducts on-demand reviews of tasks and specifications without running pipelines.
