Skill engine: weighted scoring with negative patterns and context guards (#12)

* Skill engine: weighted scoring, negative patterns, context guards, max 3 cap

Replace binary keyword matching with confidence-scored suggestions.
Phrases carry weights (0.0-1.0), negative patterns suppress false
positives, and context guards require confirming words for low-confidence
matches. Results ranked by score and capped at 3 per prompt.

* Fix CodeRabbit review: lowercase negative pattern, add missing docstring

- "DockerClient" → "dockerclient" in docker skill's negative list
  (mixed-case never matched against lowered input)
- Add docstring to main() to meet 80% coverage threshold

---------

Co-authored-by: AnExiledDev <AnExiledDev@users.noreply.github.com>

Author: AnExiledDev

.devcontainer/CHANGELOG.md

@@ -19,6 +19,13 @@
 
 ### Changed
 
+#### Skill Engine: Auto-Suggestion
+- **Weighted scoring** — Skill suggestion phrases now carry confidence weights (0.0–1.0) instead of binary match/no-match. Specific phrases like "build a fastapi app" score 1.0; ambiguous phrases like "start building" score 0.2
+- **Negative patterns** — Skills can define substrings that instantly disqualify them. Prevents `fastapi` from triggering when discussing `pydantic-ai`, and `docker` from triggering for `docker-py` prompts
+- **Context guards** — Low-confidence matches (score < 0.6) require a confirming context word elsewhere in the prompt. "health check" only suggests `docker` if "docker", "container", or "compose" also appears
+- **Ranked results, capped at 3** — Suggestions are sorted by score (then priority tier), and only the top 3 are returned. Eliminates 6+ skill suggestion floods
+- **Priority tiers** — Explicit commands (priority 10) outrank technology skills (7), which outrank patterns (5) and generic skills (3) when scores tie
+
 #### Claude Code Installation
 - **Claude Code now installs as a native binary** — uses Anthropic's official installer (`https://claude.ai/install.sh`) via new `./features/claude-code-native` feature, replacing the npm-based `ghcr.io/anthropics/devcontainer-features/claude-code:1.0.5`
 - **In-session auto-updater now works without root** — native binary at `~/.local/bin/claude` is owned by the container user, so `claude update` succeeds without permission issues

.devcontainer/plugins/devs-marketplace/plugins/skill-engine/README.md

@@ -39,11 +39,14 @@ Two capabilities:
 
 ### Auto-Suggestion
 
-The `skill-suggester.py` hook matches user prompts against keyword maps defined in each skill. When a match is found, it injects a suggestion via `additionalContext` so Claude knows a relevant skill is available.
+The `skill-suggester.py` hook scores user prompts against keyword maps for each skill using weighted matching. Suggestions are ranked by confidence and capped at **3 skills maximum** per prompt.
 
-Keywords are defined per skill as:
-- **Phrases** — Multi-word patterns (e.g., "docker compose", "REST API")
-- **Terms** — Single keywords (e.g., "FastAPI", "pytest")
+Each skill defines:
+- **Phrases** — `(substring, weight)` tuples. Weight 0.0–1.0 reflects specificity (e.g., `("build a fastapi app", 1.0)` vs `("pydantic model", 0.3)`)
+- **Terms** — Whole-word regex patterns, all scored at 0.6
+- **Negative patterns** — Substrings that instantly disqualify a skill (e.g., `"pydanticai"` suppresses `fastapi`)
+- **Context guards** — Required co-occurring words for low-confidence matches. When the best score is below 0.6, at least one guard word must appear in the prompt or the match is dropped
+- **Priority** — Integer tie-breaker (10 = commands, 7 = tech, 5 = patterns, 3 = generic)
 
 ## How It Works
 
@@ -56,9 +59,12 @@ User submits a prompt
         |
         +-> skill-suggester.py
               |
-              +-> Scan prompt against all skill keyword maps
-              +-> Match found? -> Inject skill suggestion as additionalContext
-              +-> No match? -> Silent (no output)
+              +-> Check negative patterns (instant disqualify)
+              +-> Score phrases (best weight) and terms (0.6)
+              +-> Enforce context guards on low-confidence matches
+              +-> Rank by score desc, priority desc
+              +-> Return top 3 as additionalContext
+              +-> No matches above threshold? -> Silent (no output)
 ```
 
 ### Skill Structure
@@ -127,7 +133,7 @@ skill-engine/
 +-- hooks/
 |   +-- hooks.json                   # UserPromptSubmit hook registration
 +-- scripts/
-|   +-- skill-suggester.py           # Keyword-based skill auto-suggestion
+|   +-- skill-suggester.py           # Weighted scoring skill auto-suggestion
 +-- skills/
 |   +-- api-design/                  # 22 skill directories
 |   +-- ast-grep-patterns/