cmsfornerd/cmsfornerd2/google-deep-research

Skill: Google Deep Research & Search

Overview

Enables Google Jules and autonomous AI agents to perform multi-step, iterative, and deep web research to gather comprehensive intelligence, verify technical facts, and synthesize detailed, context-ready Open Knowledge Format (OKF) v0.2 Markdown documents.

This skill integrates the /learn (or "to learn") Knowledge Ingestion Protocol to intercept external URLs, technical specifications, and raw documentation, transforming them into structured, trust-verified OKF v0.2 assets using deterministic uv python execution to minimise AI token usage.

Context Integration & File Capabilities

• Input Files: Ingests repository files (codebases, .md files, internal spec docs) or attached external documents (PDF, TXT, CSV) to cross-reference or ground web research within project-specific data.
• Output Files: Generates structured Markdown (.md) reports, system documentation, or configuration files directly into the project's repository workspace for final review in Pull Requests.

Knowledge Ingestion Protocol (/learn or "to learn")

Whenever a task or prompt includes /learn [Topic, URL, or Context] or "to learn":

1. Intercept & Parse: Suspend standard conversational behavior. Parse the target URL, raw text, or technical spec.
2. Transform to OKF v0.2: Extract core technical truths and generate/update .md documents with the Five Trust & Freshness Pillars:

• spec_version: Explicitly set to "0.2".
• type: Functional classification ("tutorial", "how-to", "reference", "explanation", "skill").
• title: Concise title string.
• status: Lifecycle state ("stable", "draft", "deprecated").
• stale_after: ISO date string (YYYY-MM-DD), typically set to 6-12 months out.
• sources: Structural list (id, title, author, url).
• generated: Provenance metadata (by, timestamp).

1. Diátaxis & UK English Alignment: Structure prose according to Diátaxis quadrants and adhere strictly to UK English spelling conventions (optimise, organise, behaviour).
2. Token Optimisation via `uv python`: Offload scanning, metadata parsing, and validation to deterministic uv python scripts (tools/migrate_okf_v02.py) to eliminate unnecessary LLM context consumption.

Workspace Quality Enforcement (markdownlint-cli)

Before presenting generated knowledge assets or finalizing file writes:

1. Programmatically invoke npx --no-install markdownlint-cli --fix <filename.md> or run script-level linting.
2. Ensure compliance with .markdownlint.json.
3. Verify frontmatter starts cleanly on Line 1, Column 1 with ---.

Deep State of Mind (DSOM) For My AI Protocol | Harisfazillah Jamel (LinuxMalaysia) | 2026-09-06 Standard: UK English | DBP-standard Bahasa Melayu Malaysia (Piawai) | GNU General Public License v3.0