New release: added README.md and example prompts to default configuration

README.md

@@ -0,0 +1,60 @@
+# WWCompanion
+
+WWCompanion is a manual, user-triggered browser extension for
+LLM-enabled AI companionship with WaterlooWorks job postings.
+
+## Table of Contents
+
+- [Usage](#usage)
+- [Configuration](#configuration)
+- [Example prompts](#example-prompts)
+  - [Example system prompt (default if not configured)](#example-system-prompt-default-if-not-configured)
+  - [Example task prompt: Generic Fit](#example-task-prompt-generic-fit)
+  - [Example task prompt: Ratings Only](#example-task-prompt-ratings-only)
+
+## Usage
+
+- Configure API settings, prompts, and tasks in **Settings**.
+- Open a WaterlooWorks job posting modal.
+- Click **Extract** or **Extract & Run** in the popup.
+- Review the streamed response, then copy or clear it as needed.
+
+## Configuration
+
+- Open **Settings** from the popup.
+- Add your API key, base URL, and model (defaults target
+  OpenAI-compatible APIs).
+- Set your system prompt and resume text.
+- Customize task presets; the top task is used as the default.
+
+## Example prompts
+
+### Example system prompt (default if not configured)
+
+Concise, critical evaluation style with a short summary and emphasized
+key points.
+
+```
+You are a precise, honest assistant. Be concise and avoid inventing details, be critical about evaluations. You should put in a small summary of all the sections at the end. You should answer in no longer than 3 sections including the summary. And remember to bold or italicize key points.
+```
+
+### Example task prompt: Generic Fit
+
+Single-section fit assessment tuned for compact output without
+interview prep.
+
+```
+You should evaluate for my fit to the job. No comparisons between postings and you don't need to suggest interview prep, we'll leave those for later. a bit of tuning to your answers: please keep things more compact, a single section for the evaluation is enough, you don't need to analyze every bullet point in the posting.
+```
+
+### Example task prompt: Ratings Only
+
+Structured numeric ratings with headings and no extra commentary.
+
+```
+Give ratings out of 10 with headings and do not include any other text.
+
+1. Fit evaluation: my fit to the role.
+2. Company status: how well this company offers career development for me.
+3. Pay: use $25 CAD per hour as a baseline; rate the compensation.
+```

background.js

@@ -1,9 +1,15 @@
 const DEFAULT_TASKS = [
   {
-    id: "task-fit-summary",
-    name: "Fit Summary",
+    id: "task-generic-fit",
+    name: "Generic Fit",
     text:
-      "Summarize the role, highlight key requirements, and assess my fit using the resume. Note any gaps and what to emphasize."
+      "You should evaluate for my fit to the job. No comparisons between postings and you don't need to suggest interview prep, we'll leave those for later. a bit of tuning to your answers: please keep things more compact, a single section for the evaluation is enough, you don't need to analyze every bullet point in the posting."
+  },
+  {
+    id: "task-ratings-only",
+    name: "Ratings Only",
+    text:
+      "Give ratings out of 10 with headings and do not include any other text.\n\n1. Fit evaluation: my fit to the role.\n2. Company status: how well this company offers career development for me.\n3. Pay: use $25 CAD per hour as a baseline; rate the compensation."
   }
 ];
 
@@ -14,7 +20,7 @@ const DEFAULT_SETTINGS = {
   apiKeyPrefix: "Bearer ",
   model: "gpt-4o-mini",
   systemPrompt:
-    "You are a precise, honest assistant. Be concise, highlight uncertainties, and avoid inventing details.",
+    "You are a precise, honest assistant. Be concise and avoid inventing details, be critical about evaluations. You should put in a small summary of all the sections at the end. You should answer in no longer than 3 sections including the summary. And remember to bold or italicize key points.",
   resume: "",
   tasks: DEFAULT_TASKS,
   theme: "system"

manifest.json

@@ -1,7 +1,7 @@
 {
   "manifest_version": 3,
   "name": "WWCompanion",
-  "version": "0.2.1",
+  "version": "0.2.2",
   "description": "AI companion for WaterlooWorks job postings.",
   "permissions": ["storage", "activeTab"],
   "host_permissions": ["https://waterlooworks.uwaterloo.ca/*"],