cue-ai 0.9.2 → 0.9.4

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

@@ -0,0 +1,124 @@
+# Referral Finder
+
+## What This Does
+
+Builds a targeted outreach list for a specific role — people who can
+get your resume in front of the hiring manager before it hits the ATS.
+
+Referrals convert at 3-5x the rate of cold applications. Most candidates
+skip this step because they don't know who to reach out to or what to say.
+This agent finds the right people and writes the right message for each one.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md`.
+Read `rules/writing-rules.md`.
+
+Extract from the JD:
+- Company name
+- Role title and department
+- Seniority level
+- Key skills and tools mentioned
+
+### Step 2 — Build the target list
+
+Identify 5 types of people worth reaching out to for this specific role:
+
+**Tier 1 — Highest conversion (contact first)**
+1. Former employees of the company who now work elsewhere
+   - They know the culture, the team, and can give you an honest read
+   - They may still have relationships inside
+   - Best found on LinkedIn: search "[Company] Alumni"
+
+2. Current employees in the same department
+   - Peer-level contact, not the hiring manager
+   - Can refer you internally or at minimum answer questions about the role
+   - Search LinkedIn: "[Company] + [Department/Role Title]"
+
+**Tier 2 — Warm path**
+3. People in your network who work at the company in any role
+   - Even a weak tie is better than a cold application
+   - Search your LinkedIn 1st and 2nd connections filtered by company
+
+4. People who made the same career transition you're making
+   - Someone who moved from your current role/company type to this type of role
+   - They can tell you what the transition actually looks like and who helped them
+
+**Tier 3 — Long shot but worth it**
+5. The hiring manager (if identifiable)
+   - Only reach out directly if you have a specific, compelling reason
+   - Most effective when you can reference shared work, a specific project of theirs, or a mutual connection
+
+### Step 3 — Write outreach messages
+
+For each tier, write a message template tailored to this specific role and company.
+
+**Format for each message:**
+
+---
+**[TIER] — [Who to target]**
+**Find them:** [Specific LinkedIn search or method]
+**What to say:**
+
+[Message — 60-90 words max. No "Hi". No "I hope this message finds you well."
+Lead with a specific, genuine reason for reaching out. Ask one clear question.
+Never ask for a referral in the first message — build the relationship first.]
+
+---
+
+### Step 4 — Write the search strategy
+
+Provide a step-by-step search playbook:
+
+```
+REFERRAL SEARCH PLAYBOOK
+Company: [name]
+Role: [title]
+─────────────────────────────────────────
+
+STEP 1 — LinkedIn search (do this first, 20 minutes)
+[Specific search strings to use]
+
+STEP 2 — Check your existing network
+[Who to look for in 1st/2nd connections]
+
+STEP 3 — Alumni networks
+[Relevant schools, companies, or communities]
+
+STEP 4 — Outreach sequence
+Week 1: Contact Tier 1 targets (2-3 people)
+Week 2: Follow up + contact Tier 2 (2-3 people)
+Week 3: Follow up on warm responses
+─────────────────────────────────────────
+
+WHAT TO DO IF THEY RESPOND
+[Specific guidance on how to move the conversation forward]
+
+WHAT NOT TO DO
+- Don't ask for a referral before building any relationship
+- Don't message more than 3 people at the same company simultaneously
+- Don't send the same message to multiple people
+```
+
+### Step 5 — Save output
+
+Write to `outputs/referral-finder.md`.
+
+Tell the user:
+> Referral playbook ready. outputs/referral-finder.md
+>
+> Start with Tier 1. One warm intro from a former employee
+> is worth 20 cold applications.
+
+## ✅ What to do next
+
+```
+npm run outreach       ← write the cold message if no warm path exists
+npm run decode         ← understand the role before reaching out
+npm run fit            ← confirm this is worth your outreach effort first
+```
+
+Apply all rules from `rules/writing-rules.md` to all messages.

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

@@ -0,0 +1,96 @@
+# Salary Research Agent
+
+## What This Does
+
+Researches current market compensation for a specific role and location.
+Gives you a realistic range, a negotiation anchor, and talking points for
+the comp conversation — before you get on a screening call.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/job-description.md` for role title and company.
+Read `inputs/my-resume.md` for years of experience and background context.
+
+### Step 2 — Extract role details
+
+From the JD identify:
+- Exact role title
+- Seniority level (entry / mid / senior / director / VP)
+- Location (remote or specific city)
+- Industry
+- Company size if mentioned
+
+### Step 3 — Research compensation
+
+Use web search to find current salary data from:
+- Glassdoor (search: "[role title] salary [city] site:glassdoor.com")
+- LinkedIn Salary (search: "[role title] salary [city] site:linkedin.com/salary")
+- Levels.fyi for tech roles (search: "[role title] [company] salary site:levels.fyi")
+- Payscale (search: "[role title] salary [city] site:payscale.com")
+- Indeed Salary (search: "[role title] salary [city] site:indeed.com")
+
+Run at least 3 searches. Collect the ranges returned.
+
+### Step 4 — Calculate ranges
+
+From all data collected, determine:
+- Floor: 25th percentile (low end, likely entry-level or low-cost market)
+- Midpoint: 50th percentile (market rate for this role/location)
+- Ceiling: 75th percentile (experienced candidate or HCOL market)
+- Stretch: 90th percentile (for negotiation anchor — never open here, use as ceiling)
+
+For sales roles, also calculate:
+- Base salary range
+- OTE (On-Target Earnings) range
+- Commission structure norms for this role type
+
+### Step 5 — Assess where the user lands
+
+Based on their resume, assess where in the range they realistically sit:
+- Years of experience in this role type
+- Industry background alignment
+- Geographic market (HCOL vs MCOL vs LCOL)
+- Whether they're targeting a step up, lateral, or step down
+
+### Step 6 — Write negotiation strategy
+
+Produce a short negotiation playbook:
+
+OPENING NUMBER: [specific number — not a range]
+Why: Opening with a range anchors to the bottom. Opening with a number anchors higher.
+
+ACCEPTABLE RANGE: [floor] to [ceiling]
+
+WALK-AWAY NUMBER: [the number below which this role doesn't make sense]
+
+IF THEY COME IN BELOW YOUR FLOOR:
+[2-3 sentences on how to respond — what to say, what to ask for instead]
+
+IF THEY ASK "WHAT ARE YOU LOOKING FOR" BEFORE AN OFFER:
+"I'm focused on finding the right fit first. I'm confident we can get to a
+number that works for both sides once we get there." — then redirect.
+
+NON-SALARY LEVERS:
+List 5 things to negotiate if base salary is fixed:
+signing bonus, equity/RSUs, remote flexibility, start date, PTO, title bump,
+performance review timing, expense account, equipment budget.
+
+### Step 7 — Save output
+
+Write to `outputs/salary-research.md`.
+
+Tell the user:
+> Market rate for [role] in [location]: [floor]-[ceiling] base, [OTE if applicable]
+> Your anchor: [opening number]
+> Full research saved to outputs/salary-research.md
+
+---
+
+## ✅ What to do next
+
+```
+npm run decode        ← decode the JD — red flags, what they really want
+npm run ats           ← check your resume clears the ATS screener
+```

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

@@ -0,0 +1,109 @@
+# Send Email — Job Search Response Sender
+
+## What This Does
+
+Takes a drafted response from the inbox scan and sends it via Gmail MCP.
+Always shows you the draft before sending — never sends without confirmation.
+
+## Prerequisites
+
+Gmail MCP connected in Claude.ai settings.
+
+## Instructions for Claude Code
+
+### Step 1 — Load drafted responses
+
+Read `outputs/inbox-scan.md`. Find all DRAFTED RESPONSE sections.
+
+If the file doesn't exist or has no drafts, tell the user:
+> Run `npm run inbox` first to scan for emails and generate drafts.
+
+### Step 2 — Present options
+
+List all available drafted responses:
+
+```
+Available drafted responses:
+
+1. Re: [subject] → [recipient] ([company])
+2. Re: [subject] → [recipient] ([company])
+...
+
+Which would you like to send? (enter number, or 'all' for all HIGH priority)
+```
+
+Wait for the user's answer.
+
+### Step 3 — Show draft for approval
+
+Show the complete draft:
+
+```
+─────────────────────────────────────────────────────
+TO:      [email address]
+SUBJECT: Re: [original subject]
+─────────────────────────────────────────────────────
+[draft body]
+─────────────────────────────────────────────────────
+
+Send this? (yes / edit / skip)
+```
+
+If they say **edit**: ask what to change, update the draft, show again.
+If they say **skip**: move to next draft.
+If they say **yes**: proceed to send.
+
+### Step 4 — Send via Gmail MCP
+
+Use Gmail MCP to:
+1. Find the original thread by message ID or subject
+2. Create a draft reply to that thread
+3. Send the draft
+
+Tell the user:
+> Sent to [name] at [company]
+> Subject: Re: [subject]
+> Sent at: [timestamp]
+
+### Step 5 — Log the send
+
+Append to `outputs/inbox-scan.md`:
+
+```
+─────────────────────────────────────────────────────
+SENT: [timestamp]
+To: [recipient]
+Subject: Re: [subject]
+─────────────────────────────────────────────────────
+```
+
+### Step 6 — Next steps
+
+After all sends:
+> All responses sent.
+>
+> If any of these lead to an interview:
+>   npm run research    ← company + interviewer brief
+>   npm run interview   ← story bank + coached answers
+>   npm run schedule    ← add to calendar
+
+## Running
+
+```bash
+npm run send-email
+```
+
+## Tone
+
+Do not change the tone or content of drafted responses without the user
+explicitly asking for edits. The drafts were written by the inbox scanner
+with the right tone already applied.
+
+## ✅ What to do next
+
+```
+npm run research    ← if an interview is scheduled, pull the company brief
+npm run interview   ← build story bank + coached answers
+npm run schedule    ← add the interview to your calendar
+npm run mock        ← practice before the real thing
+```

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

@@ -0,0 +1,127 @@
+# Job Tracker Auto-Fill
+
+## What This Does
+
+Reads all your outputs and automatically rebuilds your job tracker with
+current data — so your pipeline is always up to date without manual copying.
+
+The job tracker in `docs/templates/JOB-TRACKER.md` never stays current
+because updating it manually is friction. This agent reads everything in
+`outputs/` and reconstructs the tracker automatically.
+
+## Instructions for Claude Code
+
+### Step 1 — Load all available outputs
+
+Read every file in `outputs/` that exists:
+- `job-shortlist.md` — companies and roles from the job search
+- `fit-score.md` — most recent fit score
+- `reality-check.md` — target tier and competitive roles
+- `company-research.md` — companies researched
+- `referral-finder.md` — companies with active referral outreach
+- `inbox-scan.md` — companies that have emailed you
+- `follow-up.md` — companies in follow-up stage
+- `interview-brief.md` — companies with scheduled interviews
+- `interview-debrief.md` — companies post-interview
+- `salary-research.md` — comp benchmarks
+- `offer-comparison.md` — active offers
+- `rejection-analysis.md` — patterns and insights
+
+Read `docs/templates/JOB-TRACKER.md` for the template format.
+
+### Step 2 — Reconstruct the pipeline
+
+Build a complete picture of where each company stands:
+
+**Stages:**
+1. RESEARCHING — in job shortlist or company research, not yet applied
+2. APPLIED — application submitted, no response
+3. SCREENING — phone screen or recruiter conversation
+4. INTERVIEWING — one or more interview rounds
+5. FINAL ROUND — last stage before decision
+6. OFFER — offer received, negotiating or deciding
+7. CLOSED — rejected, declined, or accepted
+
+For each company: infer stage from which output files mention it and
+what stage they're described at.
+
+### Step 3 — Write the updated tracker
+
+Format:
+
+```markdown
+# Job Search Tracker
+Last updated: [today's date]
+
+## Pipeline Summary
+Total active: [n] | Applied: [n] | Screening: [n] | Interviewing: [n] | Offers: [n]
+Callback rate: [n]% ([n] responses from [n] applications)
+
+─────────────────────────────────────────────────────
+
+## Active Pipeline
+
+| Company | Role | Stage | Applied | Last Contact | Next Action | Notes |
+|---|---|---|---|---|---|---|
+[one row per company, populated from outputs]
+
+─────────────────────────────────────────────────────
+
+## Offers
+[If any offers exist — full detail from salary research and comparison]
+
+─────────────────────────────────────────────────────
+
+## Closed
+| Company | Role | Outcome | Date | Notes |
+|---|---|---|---|---|
+[rejections, declines, accepted offers]
+
+─────────────────────────────────────────────────────
+
+## Search Health
+Target roles: [from reality check]
+Strong callback sources: [from rejection analysis if available]
+Market comp range: [from salary research if available]
+
+─────────────────────────────────────────────────────
+
+## Next Actions (this week)
+[5-7 specific action items pulled from all outputs — follow-ups due,
+interviews to prep for, references to brief, responses to send]
+```
+
+### Step 4 — Flag urgent items
+
+At the top of the output, list anything time-sensitive:
+
+```
+⚠️  URGENT — needs action today or tomorrow:
+- [anything with a deadline from deadline-manager output]
+- [unanswered inbox alerts from inbox-scan]
+- [follow-ups more than 5 days old]
+```
+
+### Step 5 — Save output
+
+Write to `outputs/job-tracker.md` (not the template — the live version).
+
+Also update `docs/templates/JOB-TRACKER.md` if the template format is stale.
+
+Tell the user:
+> Tracker updated. outputs/job-tracker.md
+>
+> [n] companies in pipeline. [n] items need action this week.
+> [If urgent:] ⚠️  [n] items need attention today.
+
+Run this weekly. The more outputs you have, the more accurate it gets.
+
+## ✅ What to do next
+
+```
+npm run inbox          ← scan for new recruiter emails first
+npm run follow-up      ← generate follow-up for anything stale
+npm run deadline       ← if any offers have deadlines
+```
+
+Apply all rules from `rules/writing-rules.md`.

package/resources/skills/skills/career/job-hunter/inputs/README.md

@@ -0,0 +1,26 @@
+# Inputs
+
+Paste your data here before running any tool.
+
+**These files are tracked templates, not git-ignored.** Do not commit your filled-in
+data. Keep your real resume, contact details, and other personal data in a local copy
+outside version control (or add an `inputs/.gitignore` of your own) so it never lands in a PR.
+
+## Which file goes with which tool
+
+| File | Fill in before running | What to put in it |
+|---|---|---|
+| `my-resume.md` | Most tools | Your full resume, plain text |
+| `my-linkedin.md` | `npm run linkedin` | Your LinkedIn profile — headline, about, experience, skills |
+| `job-description.md` | Most tools | The full job posting you're targeting |
+| `job-search-criteria.md` | `npm run jobs` | Role titles, location, keywords, filters |
+| `apply-linkedin-url.txt` | `npm run linkedin-scrape` | Your LinkedIn profile URL (advanced) |
+| `search-outreach-target.md` | `npm run outreach` | Target company and contact details |
+| `interview-context.md` | `npm run research` | Interviewer name, title, interview type |
+
+## Tips
+
+- Plain text is fine for resume and LinkedIn — don't worry about formatting
+- For `job-description.md`, paste the complete posting including requirements
+- One job description at a time — overwrite it for each new application
+- Your resume in `my-resume.md` is your master resume — the tools tailor it per role

package/resources/skills/skills/career/job-hunter/inputs/apply-linkedin-url.txt

@@ -0,0 +1,8 @@
+# LinkedIn Profile URL
+
+Paste your LinkedIn profile URL below — just the URL, nothing else.
+
+Example: https://www.linkedin.com/in/yourname/
+
+YOUR URL:
+

package/resources/skills/skills/career/job-hunter/inputs/interview-context.md

@@ -0,0 +1,24 @@
+# Interview Context
+
+<!-- Fill this in before running npm run research -->
+
+## Interviewer Name (if known)
+
+
+## Interviewer Title (if known)
+
+
+## Interview Type
+<!-- screening / hiring manager / panel / final round -->
+
+
+## Interview Date and Time
+
+
+## Format
+<!-- phone / video / in-person -->
+
+
+## Notes
+<!-- Anything else relevant — who referred you, any prior conversations, etc. -->
+

package/resources/skills/skills/career/job-hunter/inputs/job-description.md

@@ -0,0 +1,20 @@
+# Target Job Description
+
+<!-- Paste the full job description here -->
+<!-- Include: role title, company, responsibilities, requirements, preferred qualifications -->
+<!-- The more complete the JD, the better the tailoring -->
+
+## Role Title:
+## Company:
+## Location:
+
+## About the Role
+
+
+## Responsibilities
+
+
+## Requirements
+
+
+## Preferred Qualifications

package/resources/skills/skills/career/job-hunter/inputs/job-search-criteria.md

@@ -0,0 +1,36 @@
+# Job Search Criteria
+
+<!-- Fill this in before running npm run jobs -->
+
+## Role Titles
+<!-- List the job titles you want to search for, one per line -->
+Senior Account Executive
+Account Executive
+Enterprise Sales Manager
+
+## Locations
+<!-- City and state, or "Remote" -->
+Charlotte, NC
+Remote
+
+## Must-Have Keywords
+<!-- Skills or terms that must appear in the posting -->
+
+
+## Exclude Keywords
+<!-- Terms that disqualify a posting — e.g. "internship", "entry level" -->
+internship
+entry level
+junior
+
+## Date Posted
+<!-- How recent: 1, 3, 7, or 30 days -->
+7
+
+## Employment Type
+<!-- full-time, part-time, contract, or any -->
+full-time
+
+## Salary Minimum (optional)
+<!-- Leave blank if not filtering by salary -->
+

package/resources/skills/skills/career/job-hunter/inputs/my-linkedin.md

@@ -0,0 +1,24 @@
+# My LinkedIn Profile
+
+<!-- Paste your full LinkedIn profile here -->
+<!-- Include: Headline, About section, all Experience entries with bullets, Skills, Education -->
+<!-- Remove contact info if you prefer — the audit doesn't need it -->
+
+## Headline
+
+
+## About
+
+
+## Experience
+
+### [Most Recent Role Title] | [Company] | [Start] - [End or Present]
+
+
+### [Previous Role Title] | [Company] | [Start] - [End]
+
+
+## Skills
+
+
+## Education

package/resources/skills/skills/career/job-hunter/inputs/my-resume.md

@@ -0,0 +1,28 @@
+# My Resume
+
+<!-- Paste your base resume here in plain text / markdown -->
+<!-- This is your master resume — all experience, not tailored to any specific role -->
+<!-- The resume-tailor agent handles the tailoring -->
+
+## [Your Name]
+[City, State] | [email] | [LinkedIn URL]
+
+## Summary
+
+
+## Experience
+
+### [Most Recent Role Title] | [Company] | [Start] - [End or Present]
+-
+-
+-
+
+### [Previous Role Title] | [Company] | [Start] - [End]
+-
+-
+-
+
+## Skills
+
+
+## Education

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

@@ -0,0 +1,24 @@
+# Cold Outreach Target
+
+<!-- Fill this in before running npm run outreach -->
+
+## Target Company
+
+
+## Target Contact Name (if known)
+
+
+## Target Contact Title (if known)
+
+
+## Why This Company
+<!-- What specifically draws you to this company — not generic, something real -->
+
+
+## Mutual Connection (if any)
+<!-- Name and how you know them -->
+
+
+## Recent Company News You're Aware Of
+<!-- Any recent developments you can reference -->
+

package/resources/skills/skills/career/job-hunter/rules/README.md

@@ -0,0 +1,37 @@
+# Rules Engine
+
+`writing-rules.md` is loaded into every agent prompt before generating output.
+
+These rules exist because AI-generated job content has tells — patterns that
+signal to recruiters the writing wasn't done by a person. The rules eliminate
+those tells before output is generated.
+
+## How it works
+
+Every agent file references these rules at the top of its prompt. The server
+and Claude Code both inject the full contents of `writing-rules.md` into each
+request automatically. You don't need to do anything — it's always on.
+
+## Contributing a rule
+
+The most valuable rules come from real job searches — things you noticed
+that got a response or killed one.
+
+Format for new rules:
+```
+- No "[phrase]" — [one line reason why it kills credibility]
+```
+
+Submit via pull request to `rules/writing-rules.md` or open an issue.
+See [docs/CONTRIBUTING.md](../docs/CONTRIBUTING.md).
+
+## What's in the rules
+
+- Hard punctuation rules (no em dashes, no ellipses)
+- Forbidden opening lines ("Hi", "I am writing to express...")
+- Filler phrases to delete on sight (proven track record, results-driven, etc.)
+- Bullet structure formula (action verb + what + result)
+- Tone rules (specific beats vague, sound like a person)
+- Email-specific rules
+- Cover letter structure
+- Resume bullet formula with examples