Documentation Generation

Global Documentation Generation

This skill governs the creation and maintenance of high-fidelity documentation across the Antigravity Agent Factory, ensuring alignment with SDLC phases and technical standards.

Process

The documentation generation process follows a structured path from harvesting context to artifact verification and system induction.

Step 1: Context Harvesting

Analyze the target (code, issue, or phase) to determine the required documentation type. This involves:

• Phase Detection: Checking the current SDLC phase in task.md.
• Target Mapping: Identifying if the output should be a feature walkthrough.md, a gate artifact like prd.md, or a system-level README.md.
• Constraint Resolution: Extracting specific constraints from the Plane issue and project standards.

Step 2: SDLC Artifact Generation

Follow the mandatory mapping for each SDLC phase. Each artifact must be generated according to the factory's high-fidelity templates:

Step 3: Technical API Extraction

For code-level documentation, extract structure and docstrings:

1. Parse Python modules using ast.
2. Extract classes, functions, and Google-style docstrings.
3. Generate Markdown API references with usage examples.

Step 4: Verification & Linkage

• Run link_checker.py to ensure all cross-references are valid.
• Validate artifacts against their respective JSON schemas (if applicable).

Step 5: System-Wide Synchronization (MANDATORY)

After creating or modifying any documentation artifact, you MUST run the following synchronization suite:

1. Artifact Counts: conda run -p D:\Anaconda\envs\cursor-factory python scripts/validation/sync_artifacts.py --sync
2. README Structure: conda run -p D:\Anaconda\envs\cursor-factory python scripts/validation/validate_readme_structure.py --update
3. Version Registry: conda run -p D:\Anaconda\envs\cursor-factory python scripts/validation/sync_manifest_versions.py --sync

Best Practices

• Atomic Artifacts: Keep each document focused on one phase or component.
• Evidence-Based: Always include logs, screenshots, or terminal outputs in walkthrough.md.
• Axiomatic Alignment: Documentation must be Accurate (Truth), Minimal (Beauty), and Helpful (Love).
• Consent-Driven: For P6/P7, ensure user approval is documented.
• Creation implies Synchronization: Never finish a task without ensuring the repository manifests accurately reflect your changes.
• Version Registration: If creating a new guide, ensure it is added to VERSION_LOCATIONS in sync_manifest_versions.py if it should be tracked.

When to Use

Use this skill for any documentation task, from small README updates to full SDLC phase gate generation.

Prerequisites

• Access to .agent/knowledge/ for standards.
• Understanding of the 7-phase SDLC meta-orchestration.