diff --git a/SKILL.md b/SKILL.md
new file mode 100644
index 0000000..6c7a7dc
--- /dev/null
+++ b/SKILL.md
@@ -0,0 +1,60 @@
+---
+name: fpf-navigation
+description: Specialized workflow for navigating the massive FPF-Spec.md document without exceeding context limits. Use when answering questions or making edits to the Foundation Pattern Framework (FPF) specification document.
+---
+# FPF Specification Navigation Skill
+
+<instructions>
+You are interacting with the Foundation Pattern Framework (FPF) specification document. This file is extremely large (~1.5M tokens). **NEVER attempt to read the entire file without limits.**
+
+The canonical location for this file is assumed to be:
+`~/Documents/fpf-docs/FPF-Spec.md`
+
+Follow this strict two-phase workflow to navigate the document:
+
+### Phase 1: Load Context & Table of Contents
+When starting a new investigation into FPF, always load the introduction and Table of Contents (TOC) first to ground yourself in the available patterns.
+The TOC and intro exist entirely before "Part A - Kernel Architecture Cluster".
+
+1. Find the line number where Part A begins:
+   `grep -n "^# Part A" ~/Documents/fpf-docs/FPF-Spec.md`
+2. Use the `read_file` tool with `start_line=1` and `end_line={LINE_NUMBER_MINUS_ONE}` to ingest the preamble and TOC. (Note: tools like `read_file` or `replace` require absolute paths, so expand `~` to your home directory when using them).
+3. Analyze the TOC to identify the exact Pattern IDs (e.g., `E.10.D2`, `A.6.B`) relevant to the user's query.
+
+### Phase 2: Surgical Section Extraction
+Once you know the exact Pattern ID you need to investigate, use the following robust AWK script via `run_shell_command`. 
+
+This script dynamically detects the heading level (e.g., `##` vs `###`) of your target pattern and extracts the entire section, including all child sub-sections, stopping only when it reaches the next sibling or parent heading.
+
+*Copy and paste this exact command, replacing `TARGET_ID` with your pattern (e.g., `E.10.D2`):*
+```bash
+awk -v tgt="TARGET_ID" '
+BEGIN { flag=0; lvl=0 }
+/^#+ / {
+    match($0, /^#+/)
+    cur_lvl = RLENGTH
+    if (!flag && index($0, tgt)) {
+        flag = 1
+        lvl = cur_lvl
+        print
+        next
+    }
+    if (flag && cur_lvl <= lvl) {
+        exit
+    }
+}
+flag { print }
+' ~/Documents/fpf-docs/FPF-Spec.md
+```
+
+### Phase 3: Recursive Dependency Loading (MANDATORY)
+FPF patterns operate in a strict dependency graph. You cannot safely apply a pattern without understanding the foundational rules it relies upon.
+
+1. Once you extract your target pattern using Phase 2, look at its **"Builds on:"** list (usually located at the top of the pattern description or in the TOC).
+2. You **MUST** repeat Phase 2 to extract and read all primary dependencies listed in "Builds on:".
+3. Evaluate constraints from the bottom up. (e.g., If `E.10.D2` builds on `A.7`, any solution you propose must explicitly comply with `A.7`'s Strict Distinction rules first).
+
+### Constraints
+- Never make blind edits. If asked to modify a pattern, you must load the complete section into context using Phase 2 before using the `replace` tool.
+- If a section contains sub-sections (e.g., `### E.10.D2:1`), extracting the top-level heading (`## `) using the methods above will safely include all sub-sections.
+</instructions>
\ No newline at end of file