execute

Orchestrates task execution with QA gates, coder retries, and completion evidence.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/pandejesal/drone-nav-sar --skill execute-pandejesal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: execute
Source: https://github.com/pandejesal/drone-nav-sar/tree/main/.swarm/bundled-skills/execute
Command: npx skills add https://github.com/pandejesal/drone-nav-sar --skill execute-pandejesal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating multi-agent code execution without a defined protocol leads to skipped reviews, unverified coder self-reports, and commits that bypass quality gates. This Skill enforces a full execution protocol for MODE: EXECUTE so every task passes through scope declaration, static analysis, review, and testing before completion. ## Core Features & Use Cases - Scope and Recovery Contract: Resolves task scope via declare_scope bindings, plan files_touched, or FILE directives, with explicit recovery paths for conflicts, expired bindings, and workspace mismatches. - Sequential QA Gates: Runs diff, syntax_check, placeholder_scan, imports, lint, build_check, and pre_check_batch (lint, secretscan, sast_scan, quality_budget) with structured retry handling that always returns failures to the coder agent. - Review and Test Pipeline: Delegates to reviewer, security-reviewer, and test_engineer agents, plus regression sweeps, test-drift checks, adversarial tests, and an optional 5-member council mode. - Use Case: An architect runtime executing a multi-task implementation plan uses this protocol to dispatch coders, verify diffs actually landed, run all gates, and print a filled completion checklist before marking any task done. ## Quick Start Load the execute protocol and run MODE: EXECUTE on the current plan, declaring scope for each task and enforcing every QA gate before marking tasks complete.

Frequently Asked Questions about execute

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I handle a QA gate failure during task execution?

Return to the coder agent with a structured rejection stating the gate name, reason, and required fix, then re-declare scope with replace_existing before retrying. Never fix the code yourself or skip the gate; resume execution at the failed step.

What is declare_scope and when must it be called?

declare_scope binds the exact file list a task may touch and must be called with replace_existing before every coder delegation, including retries. Without it, scope-guard blocks all coder writes.

Does pre_check_batch replace code review?

No. pre_check_batch only runs four automated static tools: lint, secretscan, sast_scan, and quality_budget. Reviewer, security review, and test_engineer verification must still run afterward; treating it as a substitute is a process violation.

How does the SAST baseline work in this protocol?

Before the first coder delegation, sast_scan runs with capture_baseline for the current phase, storing pre-existing findings per phase. Later pre_check_batch calls diff against that baseline so only new findings cause failures.

What happens when council_mode is enabled?

Council mode replaces the standard reviewer and test_engineer steps with a parallel 5-member council of critic, reviewer, sme, test_engineer, and explorer. Verdicts are submitted via submit_council_verdicts and blocking concerns return the task to the coder.

Why verify coder output with diff after completion reports?

Coders can report DONE without producing any actual change, since large writes can silently fail. Running diff and reading at least one changed file confirms the change landed before forwarding to downstream gates.