cue-ai 0.9.2 → 0.9.4

package/resources/skills/skills/career/job-hunter/agents/apply-decode-jd.md

@@ -0,0 +1,117 @@
+# Job Description Decoder
+
+## What This Does
+
+Reads between the lines of a job description. What are they really asking for?
+What do the red flags mean? What does this role actually pay? What kind of
+manager posts like this?
+
+Most job seekers read a JD at face value. This reads it the way a recruiter
+or insider would.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/job-description.md`.
+Read `inputs/my-resume.md` for context on fit assessment.
+
+### Step 2 — Decode the JD
+
+Produce the following sections:
+
+---
+
+### WHAT THEY'RE ACTUALLY LOOKING FOR
+
+Strip away the corporate language and say what this role really needs.
+Not "5+ years of experience in enterprise sales" — but "they had someone
+leave who owned a specific segment and they need someone who can hit the ground
+running without hand-holding."
+
+2-4 sentences of plain-language translation.
+
+---
+
+### RED FLAGS
+
+Flag any of the following language patterns if present, and explain what they
+usually mean in practice:
+
+- "Fast-paced environment" — often means disorganized or understaffed
+- "Wear many hats" — often means no clear role definition and scope creep
+- "Self-starter" / "entrepreneurial" — often means minimal support or training
+- "Competitive salary" without a number — often means below market
+- "Must be comfortable with ambiguity" — often means leadership doesn't have a plan
+- "Rock star" / "ninja" / "guru" — often signals culture problems
+- "Like a family here" — often means poor boundaries and unpaid overtime expectations
+- "Unlimited PTO" — often means pressure not to take it
+- "Results-oriented" without defining results — often means undefined success metrics
+- Long list of requirements for a mid-level role — often means they've burned through people
+- Requirements that span 3 different job families — often means they want one person to do three jobs
+
+For each red flag found: quote the exact phrase, explain what it usually signals,
+and rate the concern: Minor / Worth asking about / Significant.
+
+---
+
+### WHAT THIS ROLE ACTUALLY PAYS
+
+Based on:
+- The role title and seniority level
+- The company size and industry (infer from JD if not stated)
+- The location
+- Whether a range is posted (if yes, note that posted minimums are often negotiable up 10-20%)
+
+Produce:
+- Likely base range
+- Likely OTE range (if sales/commission role)
+- What the posting language signals about comp philosophy
+  (e.g. "competitive compensation" with no number = likely underpaying vs market)
+
+---
+
+### WHAT KIND OF MANAGER POSTED THIS
+
+Based on the language, structure, and emphasis of the JD, characterize the
+likely manager or culture:
+
+- Heavy on requirements, light on what you'll get = transactional culture
+- Detailed about growth and learning = likely invests in people
+- Lots of buzzwords = likely leadership-by-trend
+- Clear metrics and outcomes = likely data-driven, good or bad
+- Vague about success metrics = likely unclear expectations
+
+Be direct. This section should help the candidate decide whether to invest
+time applying.
+
+---
+
+### FIT ASSESSMENT
+
+Based on the resume:
+- Where this candidate is strong for this role (specific)
+- Where they have gaps (specific)
+- The one thing they need to nail in the screening call to get past the JD concerns
+
+---
+
+### QUESTIONS TO ASK IN THE INTERVIEW
+
+5 questions this JD specifically warrants asking — to validate or disprove
+the red flags and get the real story on the role.
+
+---
+
+### Save output
+
+Write to `outputs/jd-decoded.md`.
+
+---
+
+## ✅ What to do next
+
+```
+npm run ats           ← check your resume against the ATS before tailoring
+npm run resume        ← tailor resume to this specific role
+```

package/resources/skills/skills/career/job-hunter/agents/apply-fit-score.md

@@ -0,0 +1,183 @@
+# Job Fit Scorer
+
+## What This Does
+
+Scores how competitive you actually are for a specific role before you apply.
+Not the ATS match score — that's software. This is the human judgment score.
+
+The ATS scanner tells you if software will filter you out.
+This tells you if a hiring manager will care.
+
+Paste your resume and a job description. Get a 1-10 fit score, an honest
+breakdown of where you're strong and where you're not, a clear Apply /
+Don't Apply / Apply With Caveats recommendation, and exactly what to address
+in the cover letter and interviews if you do apply.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md`.
+Read `rules/writing-rules.md`.
+
+If `outputs/reality-check.md` exists, read it — use the market positioning
+already established there to inform this assessment.
+
+### Step 2 — Analyze fit across five dimensions
+
+Score each dimension 1-10 and explain the score specifically.
+
+**Dimension 1 — Experience Match**
+Does the candidate have relevant experience doing the same or similar work?
+- 9-10: Almost identical role at a comparable company
+- 7-8: Closely related role, similar scope
+- 5-6: Adjacent experience, some overlap
+- 3-4: Thin overlap, significant stretch
+- 1-2: Little to no relevant experience
+
+**Dimension 2 — Level Match**
+Is the candidate at the right seniority level for this role?
+- Over-qualified, Under-qualified, or Right-qualified?
+- Over-qualified is not automatically good — companies often reject people
+  they think will leave once something better comes along
+
+**Dimension 3 — Industry / Domain Match**
+Does their background map to the industry, buyer type, or domain?
+- Selling healthcare software to hospital CFOs vs. selling retail POS to SMBs
+  are different skills even if both are "sales"
+- Be specific about the mismatch or match
+
+**Dimension 4 — Track Record Match**
+Does their quantified performance align with what this role expects?
+- If the JD implies $2M quota and their biggest was $500K — flag it
+- If the JD wants enterprise and they've done SMB — flag it
+- If their numbers are strong and directly relevant — note it
+
+**Dimension 5 — Culture / Company Stage Match**
+Does their background suggest they'd thrive in this environment?
+- Startup vs. enterprise
+- High-growth vs. stable
+- Founder-led vs. professionally managed
+- Infer from JD language and company signals
+
+### Step 3 — Fit Score Report
+
+Write the report in this exact structure:
+
+---
+
+```
+JOB FIT SCORE
+Role: [title] at [company]
+─────────────────────────────────────────────────────
+```
+
+**OVERALL FIT: [X]/10**
+
+**RECOMMENDATION: APPLY / DON'T APPLY / APPLY WITH CAVEATS**
+
+One sentence explaining the recommendation. Direct.
+
+---
+
+**SCORECARD**
+
+| Dimension | Score | Summary |
+|---|---|---|
+| Experience Match | X/10 | one line |
+| Level Match | X/10 | one line |
+| Industry / Domain | X/10 | one line |
+| Track Record | X/10 | one line |
+| Culture / Stage | X/10 | one line |
+
+---
+
+**WHERE YOU'RE STRONG FOR THIS ROLE**
+
+2-4 specific things from the resume that directly match what this role needs.
+These are your selling points — use them in the cover letter and interviews.
+
+---
+
+**WHERE YOU'RE WEAK FOR THIS ROLE**
+
+2-4 specific gaps between what the JD requires and what the resume shows.
+Be precise — not "limited experience" but "JD requires enterprise deals over $500K,
+resume shows largest deal was $180K at mid-market."
+
+---
+
+**IF YOU APPLY: HOW TO FRAME IT**
+
+If the recommendation is Apply or Apply With Caveats:
+
+- The angle — the single most compelling reason to hire this person despite the gaps
+- What to emphasize in the cover letter (2-3 specific points)
+- What to prepare to address in the screening call (the objections they'll raise)
+- What NOT to bring up or volunteer
+
+---
+
+**IF YOU DON'T APPLY: WHY NOT AND WHAT INSTEAD**
+
+If the recommendation is Don't Apply:
+
+- The specific reason this application will likely fail
+- What this role would actually require (honest gap assessment)
+- 2-3 alternative role types that would be a better use of this application effort
+- What would need to change to make this role viable in 12-18 months
+
+---
+
+### Step 4 — Save output
+
+Write to `outputs/fit-score.md`.
+
+Tell the user:
+> Fit score complete: [X]/10 — [APPLY/DON'T APPLY/APPLY WITH CAVEATS]
+> Saved to outputs/fit-score.md
+
+If the score is 7 or above, continue automatically:
+> Score is [X]/10 — running ATS scan next.
+Then execute `agents/apply-ats-scan.md` against the same resume and JD.
+
+### Step 5 — Pipeline continuation
+
+If APPLY or APPLY WITH CAVEATS and score >= 7:
+Tell the user the recommended next sequence:
+```
+npm run ats           ← check ATS issues
+npm run resume        ← tailor resume
+npm run cover-letter  ← write cover letter
+```
+
+If DON'T APPLY or score < 5:
+Tell the user:
+```
+npm run reality-check ← get a broader picture of where you're competitive
+npm run jobs          ← find roles that fit your actual profile
+```
+
+## ✅ What to do next
+
+**Score 7-10 / Apply:**
+```
+npm run ats           ← fix ATS issues before submitting
+npm run resume        ← tailor resume to this specific role
+npm run cover-letter  ← write the cover letter
+```
+
+**Score below 7 / Don't Apply:**
+```
+npm run reality-check ← understand your actual market before applying anywhere
+npm run jobs          ← find roles you're genuinely competitive for
+```
+
+## Tone
+
+Honest. A 4/10 is a 4/10. Don't soften it to a 6 to spare feelings.
+A candidate who applies to a 4/10 role wastes time they could spend on a 8/10.
+The most useful thing this tool can do is tell someone not to apply.
+
+Apply all rules from `rules/writing-rules.md` to all written content.

package/resources/skills/skills/career/job-hunter/agents/apply-linkedin-audit.md

@@ -0,0 +1,74 @@
+# LinkedIn Profile Analyzer
+
+## Instructions for Claude Code
+
+You are a senior B2B sales recruiter and LinkedIn optimization specialist. Your job is to give an honest, direct audit of a LinkedIn profile — the kind of feedback a good headhunter gives behind closed doors, not the polished version they say to your face.
+
+## Step 1 — Load inputs
+
+Read `inputs/my-linkedin.md`. This contains the user's LinkedIn profile.
+Read `inputs/job-description.md`. This contains their target role.
+Read `rules/writing-rules.md`. These rules govern all rewritten content you produce.
+
+## Step 2 — Analyze each section
+
+Audit the following sections in order. For each section give:
+- A rating: Strong / Needs Work / Critical Fix
+- 2-3 sentences of specific feedback
+- A rewritten version (if Needs Work or Critical Fix)
+
+### Headline
+- Is it buyer-facing or resume-facing? Buyers skim headlines — they need to know immediately what problem you solve, not your job title.
+- Does it contain keywords a recruiter searching for this role would use?
+- Is it specific enough to be memorable?
+
+### About Section
+- Does it open strong? The first two lines appear before "see more" — they must hook.
+- Is it written in first person and does it sound like a human?
+- Does it tell a coherent story about why this person, this role?
+- Does it contain numbers or specifics, or only vague claims?
+
+### Experience Section (most recent 2-3 roles)
+- Do the bullets lead with action verbs?
+- Do the bullets contain measurable results?
+- Are the bullets written in the language of the target role (from job-description.md)?
+- Any "responsible for" or filler phrases? Flag and rewrite them.
+
+### Skills Section
+- Are the top 3 pinned skills relevant to the target role?
+- Any obvious missing skills from the job description?
+
+### Profile Completeness
+- Custom URL set?
+- Featured section populated?
+- Recommendations present?
+
+## Step 3 — Keyword gap analysis
+
+Compare the job description to the profile. List:
+- Keywords in the JD that appear nowhere in the profile (add these)
+- Keywords in the profile that are irrelevant to the target role (consider removing)
+
+## Step 4 — Priority action list
+
+Output a numbered list of the top 5 changes to make, ordered by impact. Be direct. "Rewrite your headline" is more useful than "consider updating your headline."
+
+## Step 5 — Save output
+
+Write the full audit to `outputs/linkedin-audit.md`.
+
+## Tone
+
+Direct. Honest. Useful. Not harsh for its own sake, but do not soften a bad headline by calling it "a good start." If something is weak, say it is weak and fix it. The user needs accurate feedback, not encouragement.
+
+Apply all rules from `rules/writing-rules.md` to any rewritten content you produce.
+
+---
+
+## ✅ What to do next
+
+Make the CRITICAL fixes on LinkedIn first. Then:
+```
+npm run resume        ← tailor your resume to the role you're targeting
+npm run decode        ← understand the JD before applying
+```

package/resources/skills/skills/career/job-hunter/agents/apply-linkedin-scrape.md

@@ -0,0 +1,255 @@
+# LinkedIn MCP Scraper — Setup & Auto-Audit
+
+## What This Does
+
+Installs and configures `linkedin-mcp-server`, scrapes your LinkedIn profile by URL,
+formats it into Job Hunter's standard input format, then runs the LinkedIn audit
+automatically — no copy-paste required.
+
+**This is the advanced path.** If you just want to paste your profile text, use
+`npm run linkedin` instead.
+
+---
+
+## Prerequisites Check
+
+Before doing anything else, verify the following are installed:
+
+```bash
+python3 --version    # needs 3.8+
+uv --version         # needs uv package manager
+google-chrome --version || chromium-browser --version   # needs Chrome or Chromium
+```
+
+If `uv` is missing, install it:
+```bash
+curl -LsSf https://astral.sh/uv/install.sh | sh
+```
+
+If Chrome is missing, download from https://www.google.com/chrome/
+
+---
+
+## Instructions for Claude Code
+
+You are setting up the LinkedIn MCP scraper and running a full profile audit.
+Follow each step in order. If any step fails, report the exact error before stopping.
+
+### Step 1 — Check prerequisites
+
+Run the following and report results:
+
+```bash
+python3 --version
+uv --version
+which google-chrome || which chromium-browser || which chromium
+```
+
+If `uv` is not found, run:
+```bash
+curl -LsSf https://astral.sh/uv/install.sh | sh
+source $HOME/.cargo/env 2>/dev/null || source $HOME/.local/bin/env 2>/dev/null || true
+```
+
+If Chrome/Chromium is not found, stop and tell the user to install Chrome from
+https://www.google.com/chrome/ before continuing.
+
+### Step 2 — First-time login (only needed once)
+
+Check if a LinkedIn session already exists:
+
+```bash
+ls ~/.linkedin-mcp/profile/Default/Cookies 2>/dev/null && echo "SESSION EXISTS" || echo "NO SESSION"
+```
+
+If NO SESSION, run the login command to open a browser window:
+
+```bash
+uvx linkedin-scraper-mcp@latest --login
+```
+
+Tell the user:
+> A Chrome window has opened. Log into LinkedIn normally — use your regular
+> username and password. Once you're logged in and can see your LinkedIn feed,
+> come back here and press Enter to continue.
+
+Wait for the user to confirm they've logged in before proceeding.
+
+If SESSION EXISTS, skip this step and tell the user their session is already saved.
+
+### Step 3 — Read the LinkedIn URL
+
+Read the file `inputs/apply-linkedin-url.txt`. This should contain a single LinkedIn
+profile URL, e.g. `https://www.linkedin.com/in/yourname/`
+
+If the file is empty or missing, stop and tell the user:
+> Please create the file `inputs/apply-linkedin-url.txt` and paste your LinkedIn
+> profile URL into it (just the URL, nothing else), then run this agent again.
+
+### Step 4 — Start the MCP server
+
+Start the LinkedIn MCP server in the background:
+
+```bash
+uvx linkedin-scraper-mcp@latest --transport streamable-http \
+  --host 127.0.0.1 --port 8765 --path /mcp \
+  --user-data-dir ~/.linkedin-mcp/profile \
+  --headless true \
+  --log-level WARNING &
+
+MCP_PID=$!
+echo "MCP server PID: $MCP_PID"
+sleep 5
+```
+
+Verify it started:
+```bash
+curl -s http://127.0.0.1:8765/mcp/health 2>/dev/null && echo "SERVER UP" || echo "SERVER NOT RESPONDING"
+```
+
+If SERVER NOT RESPONDING after 10 seconds, try increasing the wait and checking again.
+If it still fails, stop and report the error output.
+
+### Step 5 — Scrape the profile
+
+Read the URL from `inputs/apply-linkedin-url.txt`.
+
+Use the MCP server at `http://127.0.0.1:8765/mcp` to call the `get_person_profile`
+tool with:
+- `profile_url`: the URL from the file
+- `sections`: `["main", "experience", "education", "skills", "contact_info"]`
+
+If the scrape returns a captcha challenge or login error:
+> LinkedIn is asking for verification. Run `uvx linkedin-scraper-mcp@latest --login`
+> to re-authenticate, then run this agent again.
+
+### Step 6 — Format into Job Hunter input
+
+Take the scraped profile data and write it to `inputs/my-linkedin.md` using this
+exact structure:
+
+```markdown
+# My LinkedIn Profile
+
+## Headline
+[scraped headline]
+
+## About
+[scraped about/summary section]
+
+## Experience
+
+### [Job Title] | [Company] | [Start Date] - [End Date or Present]
+[bullet points from description]
+
+[repeat for each role]
+
+## Skills
+[comma-separated list of top skills]
+
+## Education
+### [Degree] | [School] | [Year]
+```
+
+Tell the user:
+> Profile scraped successfully. Saved to inputs/my-linkedin.md — here's a preview
+> of the first few lines:
+
+Show the first 20 lines of `inputs/my-linkedin.md`.
+
+### Step 7 — Shut down the MCP server
+
+```bash
+kill $MCP_PID 2>/dev/null || pkill -f "linkedin-scraper-mcp" 2>/dev/null
+echo "MCP server stopped"
+```
+
+### Step 8 — Check for job description
+
+Check if `inputs/job-description.md` has been filled in (more than 200 characters
+of non-template content).
+
+If it's empty or still a template, tell the user:
+> Your LinkedIn profile has been scraped and saved. Before running the audit,
+> paste a job description into `inputs/job-description.md`, then run:
+> `npm run linkedin`
+
+If it's filled in, proceed directly to Step 9.
+
+### Step 9 — Run the LinkedIn audit
+
+Read `inputs/my-linkedin.md` (just scraped).
+Read `inputs/job-description.md`.
+Read `rules/writing-rules.md`.
+
+Now perform the full LinkedIn audit as defined in `agents/linkedin-analyzer.md`.
+
+Execute every step in that agent and write the output to `outputs/linkedin-audit.md`.
+
+Tell the user:
+> Audit complete. Your results are in outputs/linkedin-audit.md
+
+Show the Priority Action List section of the output so the user sees their top
+5 actions immediately.
+
+---
+
+## Running This Agent
+
+```bash
+# First time — reads URL, logs in, scrapes, audits
+claude "follow agents/linkedin-scraper-setup.md"
+
+# After first login — just scrape and audit
+claude "follow agents/linkedin-scraper-setup.md"
+```
+
+Or with npm:
+```bash
+npm run linkedin-scrape
+```
+
+---
+
+## Troubleshooting
+
+**Captcha / login errors**
+```bash
+uvx linkedin-scraper-mcp@latest --login
+```
+Log in manually in the browser window, then re-run the agent.
+
+**Server won't start**
+```bash
+pkill -f linkedin-scraper-mcp
+uvx linkedin-scraper-mcp@latest --transport streamable-http --host 127.0.0.1 --port 8765 --path /mcp --headless true
+```
+
+**Profile not found / empty sections**
+Some profiles are private or have restricted visibility. The scraper can only
+access what LinkedIn shows to logged-in users. If your own profile comes back
+empty, make sure you're logged into the correct account.
+
+**uv command not found after install**
+```bash
+source $HOME/.local/bin/env
+```
+Then re-run the agent.
+
+---
+
+## Privacy Note
+
+Your LinkedIn credentials are stored locally in `~/.linkedin-mcp/profile/` —
+the same place Chrome stores any browser profile. They never leave your machine.
+The scraper runs Chrome on your computer; it does not proxy through any third-party
+service.
+
+---
+
+## ✅ What to do next
+
+Your profile is saved to `inputs/my-linkedin.md`. Run the audit:
+```
+npm run linkedin      ← full audit against your target role
+```