cue-ai 0.9.2 → 0.9.4

package/resources/skills/skills/career/job-hunter/agents/search-interview-scorecard.md

@@ -0,0 +1,164 @@
+# Your Interview Scorecard
+
+## What This Does
+
+Builds your personal scorecard for evaluating a company — the specific
+things that would make you say no to this role, before you get swept up
+in the excitement of being wanted.
+
+Candidates spend all their energy trying to impress. The companies that
+are worth your time are also trying to impress you. This agent builds
+the criteria you'll use to evaluate them, so when the offer comes you're
+deciding with your head, not your relief.
+
+Run this before the first interview. Use it to evaluate throughout the process.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md`.
+Read `outputs/company-research.md` if it exists.
+Read `outputs/salary-research.md` if it exists.
+Read `rules/writing-rules.md`.
+
+Ask the user:
+```
+A few questions to build your scorecard:
+
+1. Why are you looking? (leaving something specific, or seeking something specific)
+2. What went wrong in your last role? (this is often what you need to avoid repeating)
+3. What's non-negotiable for you right now?
+   (comp floor, remote/hybrid, manager quality, growth path, culture, stability, industry)
+4. What have you compromised on before that you regret?
+5. What does "this was the right move" look like in 2 years?
+```
+
+Wait for their answers.
+
+### Step 2 — Build the scorecard
+
+Create a personal evaluation scorecard with 6-8 criteria specific to this person.
+
+For each criterion:
+
+**[CRITERION NAME]**
+Why it matters to you: [from their answers — make it personal]
+What good looks like: [specific, not vague]
+What bad looks like: [specific signals to watch for]
+How to verify it: [the question to ask, the person to ask it to, the signal to watch for]
+Current signal from what I know: [based on JD and company research if available]
+Weight: Critical / Important / Nice-to-have
+
+---
+
+Common criteria to consider (adapt to their specific situation):
+
+**Compensation**
+Not just base — total comp, bonus structure, equity, benefits, trajectory.
+Question: "Can you walk me through the comp structure and how it progresses?"
+
+**Manager Quality**
+Your direct manager determines 80% of your day-to-day experience.
+Question: "How would you describe your management style? What do your direct reports say about working with you?"
+Signal: Do they ask you questions? Do they talk about their team's success or their own?
+
+**Growth Path**
+Is there a real path from this role or is it a terminal position?
+Question: "Where have people who held this role previously gone within the company?"
+Signal: Can they name 2-3 people?
+
+**Team Stability**
+Constant churn destroys productivity and signals dysfunction.
+Question: "How long has the current team been together? Has there been significant turnover recently?"
+Signal: How they answer, not just what they say.
+
+**Real Autonomy**
+The difference between "you'll own this" in the JD and what it actually means.
+Question: "Walk me through how a decision gets made in this role. When do I need approval?"
+Signal: Specificity of their answer.
+
+**Culture Honesty**
+Every company says "great culture." What they mean varies.
+Question: "What's something about working here that surprised you — good or bad?"
+Signal: Are they willing to name something real?
+
+**Financial Stability**
+Is this company going to be here in 2 years?
+From company research — funding runway, revenue signals, layoff history.
+
+**Work-Life Sustainability**
+Realistic expectations for hours, availability, urgency culture.
+Question: "What does a typical week look like for someone in this role?"
+Signal: "We work hard but we play hard" = red flag.
+
+### Step 3 — The go/no-go framework
+
+Write a clear decision guide:
+
+```
+YOUR GO / NO-GO FRAMEWORK
+
+MUST HAVE (any of these missing = decline):
+[List their critical criteria]
+
+STRONG PREFERENCE (2+ missing = serious concern):
+[List their important criteria]
+
+NICE TO HAVE (factor in but not deciding):
+[List their lower-weight criteria]
+
+THE ONE QUESTION THAT DECIDES IT:
+[Based on their specific situation — the single question they need answered]
+```
+
+### Step 4 — Evaluation tracker
+
+Create a scorecard table to fill in after each round:
+
+```
+EVALUATION TRACKER
+
+Criterion          | Weight    | What I learned | Score (1-5) | Questions remaining
+─────────────────────────────────────────────────────────────────────────────────────
+[criterion 1]      | Critical  |                |             |
+[criterion 2]      | Critical  |                |             |
+[criterion 3]      | Important |                |             |
+...
+─────────────────────────────────────────────────────────────────────────────────────
+OVERALL:                                        |             |
+```
+
+### Step 5 — The reality check note
+
+End with this:
+
+```
+A NOTE ON THE OFFER HIGH
+
+When an offer comes — if it comes — you will feel relieved. That relief
+will make everything look better than it is. That's normal. That's human.
+
+Before you say yes, fill in the scorecard completely.
+The criteria you set before you knew you had the offer
+are more reliable than the judgment you'll have after.
+
+If 2 of your Critical criteria are unfilled or concerning —
+that's information. Take 24 hours before deciding.
+```
+
+### Step 6 — Save output
+
+Write to `outputs/interview-scorecard.md`.
+
+## ✅ What to do next
+
+```
+npm run panel-decode   ← understand who you're evaluating and who's evaluating you
+npm run questions      ← build your question bank to fill in your scorecard
+npm run debrief        ← update your scorecard after each round
+npm run compare        ← compare this offer against others using your criteria
+```
+
+Apply all rules from `rules/writing-rules.md`.

package/resources/skills/skills/career/job-hunter/agents/search-jobs.md

@@ -0,0 +1,149 @@
+# Job Search Agent
+
+## What This Does
+
+Searches for live job postings across Indeed, LinkedIn, Glassdoor, and other
+platforms using JobSpy MCP. Scores each posting against your resume and outputs
+a ranked shortlist with match scores and recommended next steps.
+
+## Prerequisites
+
+JobSpy MCP server installed:
+```bash
+pip install jobspy-mcp-server
+```
+(Requires Python 3.10+.)
+
+Or via Docker (recommended fallback — works on any Python version):
+```bash
+docker run -p 9423:9423 borgius/jobspy-mcp-server
+```
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-search-criteria.md`. This contains the user's search preferences.
+Read `rules/writing-rules.md`.
+
+### Step 2 — Start JobSpy MCP server
+
+First check if a server is already running:
+```bash
+curl -s http://localhost:9423/health && echo "SERVER ALREADY UP"
+```
+
+If not running, try to start locally:
+```bash
+python3 -m jobspy_mcp_server --port 9423 &
+MCP_PID=$!
+sleep 3
+curl -s http://localhost:9423/health && echo "SERVER UP" || echo "SERVER DOWN"
+```
+
+If the local server does not start, try Docker:
+```bash
+docker run -d -p 9423:9423 --name jobspy-mcp borgius/jobspy-mcp-server
+sleep 5
+curl -s http://localhost:9423/health && echo "DOCKER SERVER UP" || echo "DOCKER SERVER DOWN"
+```
+
+If neither is available, tell the user to install JobSpy MCP first (see Prerequisites)
+and stop here.
+
+### Step 3 — Parse search criteria
+
+From `inputs/job-search-criteria.md` extract:
+- Role titles to search (may be multiple)
+- Locations (city, state, or remote)
+- Keywords to include
+- Keywords to exclude
+- Date posted preference (last 7 days default)
+- Full-time vs contract preference
+
+### Step 4 — Execute searches
+
+Use the JobSpy MCP server at `http://localhost:9423` to search for jobs.
+Run one search per role title. Collect all results.
+
+For each search, use these parameters:
+- search_term: role title
+- location: from criteria
+- results_wanted: 20
+- hours_old: 168 (7 days)
+- site_name: ["indeed", "linkedin", "glassdoor"]
+
+### Step 5 — Score each posting against resume
+
+For each job posting returned, score it 1-10 against the resume on:
+- Title match (does the role title align with the user's background?)
+- Keyword overlap (how many JD keywords appear in the resume?)
+- Experience level match (seniority alignment)
+- Industry match (does the company industry align with their background?)
+
+Discard any postings scoring below 5.
+
+### Step 6 — Output ranked shortlist
+
+Write `outputs/job-shortlist.md` with:
+
+```
+JOB SEARCH RESULTS
+Searched: [date]
+Criteria: [role] in [location]
+Total found: [n] | After scoring: [n]
+
+─────────────────────────────────────
+RANK 1 — SCORE: 9/10
+Role: [title]
+Company: [company]
+Location: [location]
+Posted: [date]
+Source: [indeed/linkedin/glassdoor]
+URL: [link]
+
+WHY IT SCORES HIGH:
+- [specific reason from resume match]
+- [specific reason]
+
+GAPS TO PREPARE FOR:
+- [requirement in JD not in resume]
+
+RECOMMENDED ACTION: Apply now / Research first / Skip
+─────────────────────────────────────
+[repeat for each posting, ranked by score]
+```
+
+### Step 7 — Stop MCP server
+```bash
+kill $MCP_PID 2>/dev/null || pkill -f jobspy_mcp_server 2>/dev/null
+docker stop jobspy-mcp 2>/dev/null && docker rm jobspy-mcp 2>/dev/null
+```
+
+### Step 8 — Prompt next steps
+
+Tell the user:
+> Found [n] matches. Top result: [role] at [company] — score [x]/10.
+> Your shortlist is saved to outputs/job-shortlist.md
+>
+> To apply to any of these:
+> 1. Copy the job URL into inputs/job-description.md
+> 2. Run: npm run resume
+> 3. Run: npm run cover-letter
+> 4. Run: npm run ats
+
+## Running
+
+```bash
+npm run jobs
+```
+
+---
+
+## ✅ What to do next
+
+```
+npm run decode    ← decode the top posting before applying
+npm run salary    ← research comp for the role before any screening call
+```

package/resources/skills/skills/career/job-hunter/agents/search-momentum-check.md

@@ -0,0 +1,194 @@
+# Weekly Momentum Check
+
+## What This Does
+
+Your weekly check-in. Measures what's actually working, flags what isn't,
+and tells you honestly whether you're making progress — or burning out.
+
+Job searching feels like failure by default because rejection is the
+default response. This agent replaces that feeling with evidence. Your
+callback rate, your funnel metrics, your patterns — measured against what's
+actually normal so you know where you stand, not where anxiety says you stand.
+
+Run this every Friday. It takes 10 minutes and changes how the weekend feels.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `outputs/job-tracker.md` if it exists.
+Read `outputs/rejection-analysis.md` if it exists.
+Read `rules/writing-rules.md`.
+
+Ask the user:
+```
+Quick rundown of your week — don't overthink it, just tell me:
+
+1. Applications submitted this week (number + types: targeted / cold ATS / referral)
+2. Responses received (callbacks, rejections, anything)
+3. Interviews or conversations that happened
+4. Anything that moved forward (second rounds, final rounds, offers)
+5. How many hours did you spend on the search this week?
+6. Honest: how are you feeling right now? (1 word or a sentence — either is fine)
+```
+
+Wait for their answer.
+
+### Step 2 — Calculate the real metrics
+
+**Callback rate this week:**
+Responses ÷ Applications submitted (this week)
+
+**Cumulative callback rate (if tracker data available):**
+Total responses ÷ Total applications
+
+**Industry benchmarks:**
+- Cold ATS applications: 2-8% callback rate is normal
+- Targeted applications (tailored + researched): 10-20% is good
+- Referral applications: 30-50% callback rate
+- LinkedIn Easy Apply: 1-3% (if it's your primary method, this is why)
+
+Compare their actual rate to benchmark. Be honest about what it means.
+
+**Funnel health:**
+- Applied → Response: [their rate] vs. benchmark
+- Response → Interview: [if data available]
+- Interview → Offer: [if data available]
+
+### Step 3 — The honest assessment
+
+Write a plain, honest paragraph. Not cheerful. Not crushing. True.
+
+If they're performing well:
+Say so specifically. "Your targeted application rate is 15%. That's above
+average. The referral you got for [company] converting to a screen means
+the referral strategy is working. Keep doing that."
+
+If they're struggling:
+Name the specific problem. "Your cold ATS callback rate is 2%. That's
+consistent with industry average for cold apps, which means the method
+is the problem, not the resume. Shifting 50% of your effort to referrals
+would likely double your callback rate."
+
+If the data is too thin to say anything useful:
+"You have 8 applications total. That's not enough data to find a pattern.
+The goal is 15-20 targeted applications before drawing any conclusions
+about what's working."
+
+### Step 4 — The week's wins
+
+Write this section no matter what the metrics say.
+
+Pull out specifically what went right this week — not vague encouragement,
+evidence of real progress:
+
+```
+THIS WEEK'S ACTUAL WINS
+
+[List 2-5 specific things — they may feel small but they are real:]
+
+- Got a callback from [company type] = your resume is landing with this audience
+- Made it through a phone screen = you can get past the first filter
+- Found a warm intro path = you have an application that won't go into a black hole
+- Completed your mock interview = you're more prepared than you were Monday
+- Sent a follow-up = most candidates don't do this
+- Identified a pattern in your rejections = that's intelligence, not failure
+- Applied to one fewer spray-and-pray role = that's strategy
+```
+
+The goal is not to minimize hard things. It's to make real things visible
+that anxiety makes invisible.
+
+### Step 5 — The burnout check
+
+Based on their hours and how they said they're feeling:
+
+**If they're overworking (8+ hours/day on the search):**
+```
+YOU ARE DOING TOO MUCH
+
+More applications does not mean more offers. The research is clear:
+quality beats quantity by a significant margin. 10 targeted applications
+outperform 100 cold ones.
+
+Next week: cap your search at 4 hours/day. Use the rest for things that
+restore you. You will not perform well in interviews when you're depleted.
+```
+
+**If they sound burnt out:**
+```
+THIS IS A REAL THING
+
+76% of job seekers experience significant stress. 45% report burnout symptoms.
+You are not unusual. You are not weak. You are human doing a hard thing.
+
+Two things that are true simultaneously:
+1. The search requires consistent effort
+2. You cannot sustain consistent effort without rest
+
+This week: take one full day off from the search. No applications,
+no LinkedIn, no email checking. You will come back sharper.
+The roles will still be there.
+```
+
+**If they're pacing well:**
+```
+YOUR PACING IS SUSTAINABLE
+
+You're putting in enough effort to make progress without burning out.
+That's harder than it sounds. Keep the rhythm.
+```
+
+### Step 6 — Next week's focus
+
+Write 3 specific actions for next week — not a long list, just 3:
+
+```
+NEXT WEEK: THREE THINGS
+
+1. [Most important — based on what's working or what needs to change]
+2. [Second priority]
+3. [One thing to protect your energy or momentum]
+```
+
+### Step 7 — The closing note
+
+Always end the momentum check with this (vary the wording, keep the message):
+
+```
+ONE MORE THING
+
+A job search is not a performance review of your worth.
+
+The hiring process has structural barriers, random timing, and biases
+that have nothing to do with your competence. Getting filtered by an
+AI screener is not a verdict on your value. Getting ghosted by a
+recruiter is not a measure of your potential. Getting rejected after
+a final round is not proof that you're not good enough.
+
+It is a process with noise in it. Your job is to navigate it with
+strategy and patience, not to internalize every outcome as signal
+about who you are.
+
+You are making progress. It doesn't always feel like it.
+Keep going.
+```
+
+### Step 8 — Save output
+
+Write to `outputs/momentum-check.md`.
+Append to the file — don't overwrite — so you build a weekly record.
+
+Tell the user:
+> Momentum check complete. Week of [date] logged.
+> [1-line summary of their actual status]
+
+## ✅ What to do next
+
+```
+npm run tracker        ← update your pipeline with this week's activity
+npm run rejection      ← if callback rate is consistently below 5%
+npm run referrals      ← if most applications are cold ATS
+```
+
+Apply all rules from `rules/writing-rules.md`.

package/resources/skills/skills/career/job-hunter/agents/search-outreach.md

@@ -0,0 +1,85 @@
+# Recruiter Cold Outreach
+
+## What This Does
+
+Writes a cold message to a recruiter or hiring manager at a company you want
+to work at — even when there's no open role posted. Harder than a cover letter
+because there's no job description to anchor to. Most people either never send
+these or send generic ones that get ignored.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md` for background and experience.
+Read `inputs/search-outreach-target.md` for target company and contact details.
+Read `rules/writing-rules.md`.
+
+### Step 2 — Research the target
+
+Use web search to find:
+- What the company does and who they sell to / serve
+- Recent news, funding, launches, or expansions (last 90 days)
+- The contact's LinkedIn title and background (if name provided)
+- Any job postings at the company that signal what teams are growing
+  (even if not the exact role — hiring in sales means the sales team is growing)
+
+### Step 3 — Find the angle
+
+The angle is the single most specific reason this person should respond.
+It must be one of:
+- A mutual connection (strongest — always lead with this if it exists)
+- A specific recent company event that connects to the candidate's background
+- A specific problem the company likely has that the candidate has solved before
+- A genuine specific reason for wanting this company over others
+
+Generic angles ("I admire your company's mission") are not angles. They are
+the reason cold messages get deleted.
+
+### Step 4 — Write three versions
+
+**Version 1 — LinkedIn message (300 characters max)**
+LinkedIn has a character limit. This version is brutally short.
+One sentence on who you are, one sentence on why you're reaching out, one ask.
+No filler. No "I hope this message finds you well."
+
+**Version 2 — Email (under 100 words)**
+Slightly more room but still disciplined.
+- Line 1: The angle — something specific, not generic
+- Line 2-3: Who you are in one or two sentences with one number
+- Line 4: The ask — specific and low-friction ("15 minutes" not "a meeting")
+- No subject line clichés: not "Exploring Opportunities" not "Quick Question"
+  Subject line should be specific: "Ex-HealthStream AE — interested in [Company]'s expansion into Southeast"
+
+**Version 3 — Referral message (if a mutual connection exists)**
+Shorter and warmer. Reference the mutual connection in line one.
+Ask the mutual to forward or make an intro — not the target directly.
+
+### Step 5 — What not to do
+
+Flag any of the following if they appear in the drafts and rewrite:
+- "I came across your profile and was impressed..."
+- "I would love to connect..."
+- "I am passionate about..."
+- "I believe I would be a great fit..."
+- Attaching a resume in a cold message (never do this — ask if they want it)
+- Asking for a job in a cold message (ask for a conversation, not a job)
+
+### Step 6 — Save output
+
+Write all three versions to `outputs/cold-outreach.md` with context on when
+to use each.
+
+## Tone
+
+The best cold messages read like they were written by a confident person who
+did their homework — not a desperate job seeker. The goal is curiosity, not sympathy.
+
+---
+
+## ✅ What to do next
+
+Send the message. Then:
+```
+npm run research      ← when they respond and book a call, prep the company brief
+```