doc-tasks-autopilot_skill

Overview

This skill automates generation and validation of implementation TASKS from SPEC or TSPEC documents, producing AI-structured TODOs with CODE-Ready scoring. It detects input type, decides whether to generate or review TASKS, creates dependency graphs, assigns priorities, and outputs validated TASKS ready for code implementation. The pipeline includes iterative review and auto-fix cycles to reach target quality thresholds.

How this skill works

The autopilot parses input identifiers (SPEC-NN, TSPEC-NN, or TASKS-NN) to determine Generate or Review mode, checks for existing TASKS files, and extracts implementation units and test requirements. It decomposes work into tasks with element IDs, priorities, upstream/downstream links, and traceability tags, runs validations (TASKS-Ready and CODE-Ready), and iterates fixes via reviewer/fixer until quality targets are met. Final output is a TASKS document with dependency graph, implementation contracts if required, and a CODE-Ready score.

When to use it

• You have a SPEC or TSPEC and need a developer-ready task breakdown for implementation.
• You need to review or validate an existing TASKS document before starting coding.
• You want automatic generation of dependency graphs, priorities, and traceability tags for implementation planning.
• When you require CODE-Ready scoring to determine if tasks are ready for developers or need further refinement.

Best practices

• Provide SPEC/TSPEC inputs with consistent IDs (TYPE-NN) so detection works reliably.
• Ensure upstream SPEC/TSPEC reach TASKS-Ready thresholds (default 90%) before generation for best results.
• Keep upstream documents well-scoped to avoid overly large TASKS outputs; use chunking when processing many specs.
• Review autogenerated implementation contracts when tasks have 3+ dependencies to confirm interfaces and state machines.
• Use the configured max_iterations (default 3) and target_score (default 90) to control fix/review cycles and avoid endless loops.

Example use cases

• Generate TASKS-05 from SPEC-05 to produce implementation TODOs, dependency graph, and CODE-Ready report.
• Review TASKS-02 in Review Mode to get a quality report with auto-fixable recommendations and warnings.
• Run a multi-spec batch: generate TASKS for SPEC-01,SPEC-02 and validate combined traceability tags and dependency DAGs.
• Invoke the fix cycle after a failed review to automatically normalize element IDs, regenerate Mermaid diagrams, and add missing sections.