cue-ai 0.9.2 → 0.9.4

package/resources/skills/skills/career/job-hunter/agents/apply-skills-gap-filler.md

@@ -0,0 +1,109 @@
+# Skills Gap Filler
+
+## What This Does
+
+Reads a job description and tells you exactly which "requirements" are real
+gates vs. recruiter wishlist — and for the real gaps, gives you the fastest
+path to closing them.
+
+Recruiters copy-paste requirements from previous job postings, add
+everything the last person in the role happened to know, and then use
+the list to filter out qualified people. Most postings have 10-12
+requirements. You need 60-70% to be competitive. This agent tells you
+which 60-70% matters and what to do about the rest.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md`.
+Read `rules/writing-rules.md`.
+
+### Step 2 — Classify every requirement
+
+Extract every stated requirement from the JD. For each one, classify it:
+
+**HARD GATE** — You will be screened out without this. Non-negotiable.
+- Specific license or certification (Series 7, RN, PE, Bar)
+- Minimum years experience explicitly stated ("5+ years required")
+- Specific technology that is core to the role (not "preferred")
+- Location or legal requirement (must be US citizen, must be on-site)
+
+**REAL REQUIREMENT** — Hiring manager will test for this directly.
+- Primary skill of the role
+- Tools they use daily
+- Domain knowledge central to the function
+
+**NICE-TO-HAVE** — Listed to describe an ideal candidate, not a gate.
+- "Preferred" or "plus" language
+- Secondary tools ("experience with Salesforce a plus")
+- Soft skills (everyone lists these, no one actually screens for them)
+- Advanced degrees when not explicitly required
+- Industry experience when transferable experience exists
+
+**INFLATED REQUIREMENT** — Sounds hard, actually learnable in weeks.
+- Common when JD was written by someone who doesn't do the job
+- Often: specific software version, methodology names, tool names
+- Tells you what to learn before the interview, not what blocks you
+
+### Step 3 — Gap analysis against the resume
+
+For each HARD GATE and REAL REQUIREMENT the candidate doesn't have:
+
+**Gap:** [what's missing]
+**Type:** Hard Gate / Real Requirement
+**Honest assessment:** Is this actually required or are they describing their last hire?
+**Time to close:** [realistic estimate — days, weeks, months]
+**How to close it:**
+- Free: [specific resource — YouTube channel, documentation, free course]
+- Fast: [certification or course that signals competency, with cost and time]
+- Deep: [if this requires significant work — be honest about the timeline]
+**What to say in the interview if you don't have it yet:**
+[Exact language for addressing the gap honestly without disqualifying yourself]
+
+### Step 4 — The Apply Decision
+
+Based on the gap analysis:
+
+**APPLY NOW** — You meet the hard gates, have the real requirements, gaps are closeable
+**APPLY + CLOSE GAPS FAST** — You meet most requirements, specific gaps closeable before interview rounds
+**APPLY IN [X WEEKS]** — You have a gap that would disqualify but can close it in a specific timeframe
+**DON'T APPLY YET** — A hard gate you cannot realistically close soon
+
+Be direct. If the honest answer is "don't apply yet because X," say that.
+
+### Step 5 — The Fast Track
+
+If the recommendation is APPLY + CLOSE GAPS FAST, write a specific
+3-7 day sprint plan:
+
+```
+GAP-CLOSING SPRINT
+
+Goal: Close [gap] before the interview stage
+
+Day 1-2: [specific resource + what to complete]
+Day 3-4: [specific resource + what to complete]
+Day 5-7: [hands-on practice + what to build or document]
+
+By end of sprint:
+- You can speak credibly about [topic]
+- You have [specific artifact/project/cert] to point to
+- Interview answer: "[exact phrasing]"
+```
+
+### Step 6 — Save output
+
+Write to `outputs/skills-gap.md`.
+
+## ✅ What to do next
+
+```
+npm run fit            ← score overall fit now that you know the gaps
+npm run resume         ← tailor resume to emphasize what you DO have
+npm run ats            ← ensure keywords from real requirements are in your resume
+npm run interview      ← prep for how to handle gap questions in the interview
+```
+
+Apply all rules from `rules/writing-rules.md`.

package/resources/skills/skills/career/job-hunter/agents/career-internal.md

@@ -0,0 +1,94 @@
+# Internal Job Application
+
+## What This Does
+
+Applies for an internal role differently than an external one. You know people,
+you have context, you have a reputation — and you have more to lose if it goes
+sideways. This agent accounts for all of that.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md` for the internal role.
+Read `rules/writing-rules.md`.
+
+### Step 2 — Gather internal context
+
+Ask:
+> A few things I need to know about the internal situation:
+> 1. Does your current manager know you're applying?
+> 2. Do you have a relationship with the hiring manager for this role?
+> 3. Who else might be applying internally?
+> 4. What's the general perception of you in the company — strong performer, under the radar, known for specific things?
+> 5. Why are you moving — growth, better fit, comp, or getting out of a bad situation? (be honest, this affects strategy)
+> 6. What would your current manager say about you if the hiring manager asked?
+
+### Step 3 — Strategy assessment
+
+Based on their answers, assess:
+
+**Visibility advantage** — what does this person's internal reputation add to
+their application that an external candidate can't match?
+
+**Political considerations** — is there anything that needs to be handled before
+or during the application? (informing current manager, managing relationships)
+
+**The real pitch** — internal applications should reference institutional knowledge,
+existing relationships, and demonstrated cultural fit. External candidates can't do this.
+
+### Step 4 — Internal resume tailoring
+
+Tailor the resume for this role with internal framing:
+- Lead experience bullets with company-specific impact where possible
+- Reference internal initiatives, products, or teams by name where relevant
+- Quantify with internal metrics the hiring manager will recognize
+
+### Step 5 — Internal cover letter
+
+Write an internal cover letter (shorter than external — 150-200 words):
+- Opens differently than an external letter — no need to introduce yourself
+- References the internal context: "Having spent [X] years building [thing] here..."
+- Names something specific about why this role is the right next move internally
+- Does not apologize for applying or over-explain the move
+- Ends with confidence — you know this company, you know what you're getting into
+
+### Step 6 — The conversation to have first
+
+Before submitting: who should the user talk to before the application goes in?
+
+Write a short script for:
+1. The conversation with their current manager (if not already had)
+2. A conversation with the hiring manager if they have a relationship
+   ("I wanted to let you know I'm planning to apply for the [role]...")
+
+These conversations, done right, are often more important than the application itself.
+
+### Step 7 — What's different about the interview
+
+Flag 3 things that are different about an internal interview:
+- They already know your reputation — the interview validates it or contradicts it
+- You can reference internal context that external candidates can't
+- The stakes are higher if you don't get it — have a plan for that scenario
+
+### Step 8 — If you don't get it
+
+Write 2-3 sentences on how to handle not getting the role with your reputation intact:
+What to say to the hiring manager, what to say to your current manager, and
+how to use the feedback constructively.
+
+### Step 9 — Save output
+
+Write to `outputs/internal-application.md`.
+
+---
+
+## ✅ What to do next
+
+Have the conversation with your current manager before submitting — if you haven't already.
+Then submit. Then prepare as you would for any interview:
+```
+npm run research      ← pull context on the team and hiring manager
+npm run interview     ← build your prep guide
+```

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

@@ -0,0 +1,173 @@
+# LinkedIn Content Generator
+
+## What This Does
+
+Writes 5 LinkedIn posts built around your real experience and target
+positioning — posts that build credibility with the people who hire for
+the roles you want.
+
+Posting on LinkedIn during a job search increases inbound recruiter contact
+meaningfully. But most job seeker LinkedIn content is either too generic
+("5 lessons from my career") or too obviously desperate ("open to work —
+DM me"). Neither builds authority.
+
+This agent writes posts that demonstrate expertise, not availability. The
+goal is to become recognizable to the right people before you need them.
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md`.
+Read `inputs/job-description.md` if available — use target role as positioning lens.
+Read `rules/writing-rules.md`.
+
+Ask the user:
+
+```
+A few questions to write posts that sound like you:
+
+1. What's the most counterintuitive thing you've learned in your career?
+   Something most people in your field get wrong.
+
+2. What's a story from your work that you find yourself telling a lot?
+   The one you use when explaining what you do.
+
+3. What do you know about [your industry/function] that took you years
+   to figure out but seems obvious now?
+
+4. What are you genuinely interested in right now — a trend, a problem,
+   a debate in your field?
+
+5. What's something you did that failed, and what did you actually learn?
+   (not the polished failure story — the real one)
+```
+
+### Step 2 — Generate 5 posts
+
+One post per type. Each one is based on their answers — not generic.
+
+---
+
+**POST TYPE 1 — The Counterintuitive Take**
+
+Format: Lead with the counterintuitive claim. Explain why most people
+think the opposite. 3-4 sentences of evidence or reasoning. One
+practical implication.
+
+Length: 150-200 words
+Hook: Starts with the claim, not setup
+Tone: Direct. Confident. Not preachy.
+No: "In my experience..." "I've learned that..." "Here's the thing:"
+
+---
+
+**POST TYPE 2 — The Story Post**
+
+Format: Open in the middle of the story (not "once upon a time").
+Build to the insight. The lesson comes last — and it's one sentence.
+
+Length: 200-250 words
+Hook: First line is a moment, not a setup
+Tone: Specific. Human. The reader should feel like they were there.
+No: Emojis as bullets. Ellipsis cliffhangers. Manufactured suspense.
+
+---
+
+**POST TYPE 3 — The Hard-Won Insight**
+
+Format: State the insight plainly. Back it with one specific example
+from their career. Three implications, stated briefly.
+
+Length: 150-175 words
+Hook: The insight, stated as a declarative sentence
+Tone: Earned authority. Not arrogant, but confident.
+No: "I used to think X but now I know Y" (overused)
+
+---
+
+**POST TYPE 4 — The Industry Take**
+
+Format: Name the debate or trend. Take a clear position.
+One supporting argument. One honest acknowledgment of the counterargument.
+Close with what you're watching.
+
+Length: 175-225 words
+Hook: The position, stated directly
+Tone: Informed. Specific to their industry/function.
+No: Fence-sitting. Generic observations. "It depends."
+
+---
+
+**POST TYPE 5 — The Useful Failure**
+
+Format: What happened (briefly). What you thought you knew.
+What actually happened. What you changed. One sentence on why it matters.
+
+Length: 200-250 words
+Hook: The moment it went wrong — specific
+Tone: Self-aware without self-flagellation. The lesson should be useful,
+not a confession.
+No: "I failed and it was the best thing that ever happened to me" framing.
+No: Vague lessons. The takeaway should be specific enough to be actionable.
+
+---
+
+### Step 3 — Posting strategy
+
+Add a one-page strategy:
+
+```
+POSTING STRATEGY
+
+Cadence: 2-3 posts per week during active search. 1 per week otherwise.
+Best days: Tuesday, Wednesday, Thursday
+Best time: 8-9am or 5-6pm in your target market's time zone
+
+Post order (recommended):
+Week 1: Counterintuitive Take (establishes POV)
+Week 2: Story Post (builds trust)
+Week 3: Hard-Won Insight (demonstrates depth)
+Week 4: Industry Take (signals relevance)
+Week 5: Useful Failure (earns credibility)
+Repeat with fresh angles.
+
+What to do after posting:
+- Reply to every comment in the first hour
+- Do not reply with "Thanks!" — add something to the conversation
+- Connect with everyone who engages who might be relevant
+
+What NOT to post:
+- Anything about being "open to work" or "exploring new opportunities"
+- Vague inspiration content: hustle culture, grind mindset, motivational quotes
+- Anything with a poll that has no real stakes
+- Job application status updates
+```
+
+### Step 4 — Profile optimization note
+
+Add a brief note:
+If their LinkedIn headline or About section doesn't match the target
+positioning, flag it and suggest they run `npm run linkedin` first.
+
+### Step 5 — Save output
+
+Write to `outputs/linkedin-content.md`.
+
+Tell the user:
+> 5 posts ready. outputs/linkedin-content.md
+>
+> Post Type 1 first — it's the fastest way to establish a POV.
+> Schedule them out, don't post all at once.
+> The goal is recognition, not virality.
+
+## ✅ What to do next
+
+```
+npm run linkedin       ← audit your profile before you start posting
+npm run network        ← warm up specific contacts once you're posting
+npm run referrals      ← leverage the visibility for warm intros
+```
+
+Apply all rules from `rules/writing-rules.md`.
+No em dashes in any post. Posts should sound like a person talking, not writing.

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

@@ -0,0 +1,262 @@
+# LinkedIn Post Scanner
+
+## What This Does
+
+Scans your LinkedIn feed for posts worth commenting on — people talking
+about job market struggles, hiring, layoffs, bad interview experiences,
+resume tips, career pivots — and drafts a genuine comment for each one.
+
+You review every comment before it goes anywhere. Nothing posts automatically.
+The automation is in the finding and drafting. The judgment is yours.
+
+One well-placed comment from the person who built a job search tool,
+on a post with 500 likes, reaches more people than a cold post of your own.
+
+## Prerequisites
+
+LinkedIn MCP connected in Claude.ai settings:
+**Settings → Integrations → LinkedIn**
+
+Or use the LinkedIn scraper MCP:
+```bash
+uvx linkedin-scraper-mcp@latest --login
+```
+
+## Instructions for Claude Code
+
+### Step 1 — Load inputs
+
+Read `inputs/my-resume.md` for your background context.
+Read `rules/writing-rules.md`.
+
+Ask the user:
+```
+A few quick questions:
+
+1. What's your LinkedIn profile URL?
+   (so the comment sounds like you specifically)
+
+2. What's the one-line description of Job Hunter you want to use?
+   (e.g. "I built a free open source job search toolkit — jobhunter.ai")
+
+3. Any topics to avoid commenting on?
+   (e.g. specific companies, political topics, anything sensitive)
+```
+
+Wait for their answers.
+
+### Step 2 — Define search targets
+
+Look for posts matching these signals — high engagement, relevant topic,
+comment window still open (under 48 hours old is ideal):
+
+**Primary targets — highest value:**
+- Posts about job market difficulty ("can't get callbacks", "100 applications")
+- Posts about recruiter behavior (ghosting, ghost jobs, bias)
+- Posts about resume and ATS struggles
+- Posts about interview processes being too long
+- Posts from recruiters giving job search advice (like the screenshot above)
+- Posts about layoffs and career pivots
+- Posts asking for resume feedback or job search advice
+
+**Secondary targets — worth engaging:**
+- Posts about LinkedIn optimization
+- Posts about salary negotiation
+- Posts about career growth and promotion
+- Posts from hiring managers explaining their process
+
+**Skip:**
+- Anything political
+- Posts from direct competitors
+- Posts where top comments are already from tools/bots (signals low quality engagement)
+- Posts older than 72 hours (engagement window mostly closed)
+- Any topic the user said to avoid
+
+### Step 3 — Scan the feed
+
+Use LinkedIn MCP to:
+1. Pull the user's home feed (last 24-48 hours)
+2. Search LinkedIn for recent posts using these queries:
+   - "job search" + high engagement filter
+   - "job market" + recent posts
+   - "resume" OR "ATS" + recruiter posts
+   - "interview process" + frustration signals
+   - "layoff" OR "laid off" + recent
+   - "hiring" + recruiter perspective
+
+For each post found, extract:
+- Author name and title
+- Post text (full)
+- Engagement count (likes + comments)
+- Posted time
+- Post URL
+
+### Step 4 — Score and filter
+
+Score each post on 3 dimensions:
+
+**Relevance (1-5):** How directly relevant is Job Hunter to this post?
+- 5: Post is exactly about the problem Job Hunter solves
+- 4: Adjacent topic where Job Hunter adds clear value
+- 3: Tangentially related
+- 1-2: Too far, skip
+
+**Reach (1-5):** How many people will see this comment?
+- 5: 500+ likes or the poster has 10k+ followers
+- 4: 200-500 likes or 5k-10k followers
+- 3: 100-200 likes
+- 1-2: Under 100, skip (not worth the effort)
+
+**Comment window (1-5):** Is there still time for the comment to get traction?
+- 5: Under 6 hours old
+- 4: 6-24 hours old
+- 3: 24-48 hours old
+- 1-2: Over 48 hours, skip
+
+**Total score: 3-15. Surface anything scoring 9+.**
+
+### Step 5 — Draft comments
+
+For each qualifying post (score 9+), draft 2 comment options:
+
+**Comment style rules:**
+- 2-4 sentences maximum — LinkedIn comments reward brevity
+- First sentence must add something specific to the conversation
+- Not "Great post!" or "This is so true" — both are noise
+- Reference something specific from the post text
+- Mention Job Hunter naturally — not as an ad, as a genuine contribution
+- Sound like a person who built something to solve this exact problem
+- Never start with "Hi" or "I" as the first word
+- No em dashes
+
+**Option A — Contribution style:**
+Add a specific insight or data point that extends the conversation,
+then mention Job Hunter as context for why you know this.
+
+*Example:*
+"The ghost job problem is worse than most people realize — postings stay up
+for months after hiring freezes because no one removes them. Built a ghost
+job detector into an open source tool (jobhunter.ai) after getting burned
+by this multiple times. Eight signals worth checking before you spend hours
+tailoring an application."
+
+**Option B — Solidarity + resource style:**
+Acknowledge the specific frustration from the post, then offer
+Job Hunter as a direct resource.
+
+*Example:*
+"100 applications, 2 callbacks — that math usually means the ATS is
+filtering before a human sees the resume. Built a free tool specifically
+for this after going through the same thing. jobhunter.ai if it's useful."
+
+### Step 6 — Present for approval
+
+Present each qualifying post and its draft comments in this format:
+
+```
+─────────────────────────────────────────────────────
+POST [n] — Score: [X]/15
+─────────────────────────────────────────────────────
+Author: [Name] — [Title]
+Posted: [X hours ago]
+Engagement: [likes] likes · [comments] comments
+URL: [url]
+
+POST SUMMARY:
+[2-3 sentence summary of what they said]
+
+OPTION A — Contribution style:
+─────────────────────
+[Draft comment]
+─────────────────────
+
+OPTION B — Solidarity + resource style:
+─────────────────────
+[Draft comment]
+─────────────────────
+
+→ Use A / Use B / Edit / Skip
+─────────────────────────────────────────────────────
+```
+
+Go through posts one at a time. Wait for the user's decision before
+showing the next one.
+
+### Step 7 — Approved comments queue
+
+Keep a running list of approved comments:
+
+```
+APPROVED COMMENTS — Ready to post
+─────────────────────────────────────────────────────
+[n] comments approved
+
+1. [Author] — [Post URL]
+   [Approved comment text]
+   [ ] Posted
+
+2. [Author] — [Post URL]
+   [Approved comment text]
+   [ ] Posted
+```
+
+### Step 8 — Post queue
+
+When the user has finished reviewing all posts, show the approved queue
+and offer two paths:
+
+**Path A — Post via LinkedIn MCP (if connected):**
+"Ready to post [n] comments. Post them now? (yes / one at a time / save for later)"
+
+If yes: post each approved comment via LinkedIn MCP to the correct thread.
+Mark each as posted.
+
+**Path B — Manual posting guide:**
+If LinkedIn MCP isn't available for posting, generate a posting guide:
+```
+MANUAL POSTING GUIDE
+
+For each comment below:
+1. Open the URL
+2. Click Comment
+3. Paste the text
+4. Post
+
+[numbered list with URL + comment text]
+```
+
+### Step 9 — Save output
+
+Write approved comments and status to `outputs/linkedin-scanner.md`.
+Append each session — don't overwrite — so you build a record.
+
+Tell the user:
+> Scan complete. [n] posts found, [n] qualifying, [n] approved.
+> [If posted:] [n] comments posted.
+> [If manual:] Manual posting guide saved to outputs/linkedin-scanner.md
+
+## Posting frequency guidance
+
+```
+SUSTAINABLE CADENCE
+
+3-5 meaningful comments per day is enough to build real visibility.
+More than 10/day starts to look like a bot — even when it isn't.
+
+Best times: 8-10am and 5-7pm in your target market's time zone.
+Best days: Tuesday through Thursday.
+
+Track which comment styles get replies — those are the ones to do more of.
+A comment that starts a real conversation is worth 10 that get ignored.
+```
+
+## ✅ What to do next
+
+```
+npm run linkedin-post  ← write original posts for your own feed
+npm run linkedin       ← audit your profile so new visitors convert
+npm run network        ← direct outreach to people who engage with your comments
+```
+
+Apply all rules from `rules/writing-rules.md`.
+No em dashes in any comment. Sound like a person.