Agents & Skills
Claude Code workflows built for sales, research, writing, and everyday use.
Each one is a plain text .md file dropped into Claude Code's config directory. The descriptions below explain what each does and how it works — enough detail to build your own version from scratch. If you'd rather start with the exact files, they're available for download.
Bible Study System
Always Free FreeAnalyzes any scripture passage from 6 simultaneous perspectives — original language, historical context, cross-references, and three life-application lenses. All agents run in parallel; outputs compile into one structured report.
- Pass any scripture reference: single verse, range, or full chapter (e.g. Genesis 3:15, Romans 12)
- You select which of the 6 sub-agents to run — or all at once
- Selected agents spin up as independent Claude instances and run in parallel
- Each returns a structured analysis in its domain
- Outputs are compiled into one formatted report and saved to a dated file
Takes a company URL and produces a full intelligence report: product, ICP, competitors, value proposition, likely objections, a real outbound plan with target accounts, and interview prep answers tailored to that specific company.
- Fetches the company website and runs ~12 targeted web searches across news, funding, Glassdoor, and job boards
- Reads 5–10 open job postings to extract every tool the company uses internally (CRM, dialer, intelligence tools) — this becomes the internal tech stack section
- Identifies 2–3 buyer personas with their specific pain language, decision process, and the metrics they're measured on
- Builds a real outbound plan: an actual target account, real people to contact, and a personalized message draft for each
- Writes verbatim answers to "why this company?" and "what do you know about our product?" based on the research
- Saves the full report to file for reference before calls or interviews
Takes a LinkedIn URL or name + company. Searches podcasts, articles, posts, and conference talks to build a pre-interview briefing: career arc, values, hiring philosophy, predicted questions with tailored answers, and conversation hooks.
- Searches for podcast appearances first — conversations reveal more than a polished bio; extracts specific quotes and recurring themes
- Analyzes LinkedIn activity: what they post about, what tone they use, what they celebrate, what they react to
- Checks for red flags: Glassdoor reviews mentioning them, any public controversies
- Synthesizes the top 5 values this person holds, ranked by evidence strength
- Predicts 5 questions based on their specific role and values, not generic interview prep — and writes tailored answers
- Identifies concrete connection points for authentic rapport, not forced small talk
Takes a company name and deep-dives every founder and key sales leader. Career history, management philosophy, Glassdoor patterns, leadership turnover signals, and red flags. Outputs a trust assessment before you commit.
- Identifies the full leadership roster from About pages, LinkedIn, and press coverage
- Runs 28+ targeted searches per leader, including explicit red flag screening: lawsuits, controversies, layoff history, comp disputes
- Analyzes Glassdoor for sales-team-specific reviews and senior management patterns
- Checks leadership tenure on LinkedIn — churn patterns tell the truth when descriptions don't
- Sales leadership gets extra scrutiny: SDR expectations, promotion timeline, quota design, coaching style
- Delivers a bottom-line trust verdict: green flags, yellow flags, red flags, and a direct assessment
Takes a list of LinkedIn URLs for people at a target company. Researches each person individually and produces 3–5 distinct conversation starter angles per contact — with example messages. One file per person.
- You drop a .txt file of LinkedIn profile URLs into the company folder and run the agent
- Runs 15+ searches per person: career history, recent posts, podcasts, projects, conferences, shared geography or background
- Looks for: content they created, career moves, shared connections, project launches, hot takes, mutual interests
- For each angle: what it is, where it was found, why it works as an opener, and an example message
- Message examples are short, specific, and casual — designed to feel like a genuine observation, not a pitch
- Cross-references with any existing company research to add context (recent funding, product launches)
Interviews you about your background and builds STAR method answers from your real experiences. Prepares your story in three lengths, objection handling frameworks, discovery call templates, and negotiation talking points.
- Asks targeted questions across your background, sales philosophy, best wins, and approach to specific scenarios
- Structures every answer using STAR (Situation, Task, Action, Result) with your actual metrics and examples
- Produces your background story in three versions: 30 seconds, 2 minutes, and 5 minutes
- Covers all major interview categories: quota attainment, prospecting, objections, why this company, compensation expectations
- Writes a list of questions you should ask interviewers, plus follow-up email templates
- Saves everything to a single prep document you can reference the day of
Monitors a list of target companies daily across news, funding data, and career pages. Classifies each signal by type and confidence, deduplicates against a 24-hour window, and delivers formatted Slack alerts. Set it once; signals arrive on a schedule.
- Target companies are configured in a YAML file — add or remove companies without touching code
- Four data sources run every cycle: NewsAPI (classifies articles as funding / leadership / product launch by keyword), Crunchbase API (structured funding rounds, 0.92 confidence), career page scraper (tracks open role count vs. stored baseline — fires when a company's hiring velocity spikes 5+ roles AND 20%+ over baseline), and LinkedIn (stubbed — architecture is in place for when scraping is feasible)
- Each signal gets a confidence score based on source reliability and specificity
- SQLite deduplication via SHA-256 (company + signal type + title) prevents the same event from firing twice within the lookback window
- Delivers Block Kit–formatted Slack messages with emoji type indicators and a run summary at the end
- Runs via APScheduler locally or Windows Task Scheduler — no cloud infrastructure needed
Interviews you section by section before writing a word. Drafts all 8 newsletter sections in your exact voice, runs a pre-publish checklist, saves the file, and stays in conversation for revisions.
- Asks questions in 6 batches of 2–3 at a time — never dumps everything at once
- Covers: issue basics and tone, the week's story, the DM breakdown (concept → wrong move → the turn → outcome → principle), the spotlight person, the tip/fact/sponsor
- Drafts all 8 sections in one pass using strict voice rules: all lowercase, 1–3 sentences per paragraph, no buzzwords, dry humor only, narrator is a practitioner learning in public
- Runs a 12-point pre-publish checklist: did "i" stay lowercase? is the tiny fact believable-but-false? does the sponsor appear twice?
- Saves to a dated draft file and asks "anything you want to adjust?" — stays in conversation until it's right
Tell it the recipe and when you need it ready. It calculates your start time, generates a clock-based cooking runsheet, flags the mistakes that are easy to make before you hit them, and answers questions live throughout the cook.
- You say /cook [recipe] ready-by [time] — e.g. /cook chili ready-by 6pm
- Reads the recipe, calculates total time, adds a 15% novice buffer, gives you an explicit start time
- Output is a runsheet organized by clock time, not step numbers: "2:30pm — BROWN THE BEEF (15 min)"
- Every step includes the exact quantity — "add 3 lbs ground beef," never "add the beef"
- Pitfall blocks appear as standalone warnings before the step where that mistake would happen
- Handles live Q&A mid-cook: "I forgot to drain the beans" → recovery plan with an updated finish time
Captures system audio, microphone, and optional screen video in the background. Transcribes live every 12 seconds during the call so you can query it mid-meeting. /stop runs full Whisper transcription on the complete recording.
- /start launches a background task capturing WASAPI loopback (whatever plays through your speakers — Zoom, Teams, browser) plus your microphone simultaneously
- Optionally records screen video from any of up to 4 monitors; you specify which ones at launch
- Live transcript updates every ~12 seconds — during a call you can ask "what did they just say about pricing?" and get an answer from the partial transcript
- Monitor switching is supported mid-session without stopping the audio recording
- /stop ends the recording, runs full Whisper transcription on the complete audio for an accurate final transcript
- All processing is local — nothing leaves your machine
Reads your full watch history, rating patterns, and hard exclusion rules. Surfaces 2–4 picks that match your taste profile. Output is title and year only — no descriptions or plot hints. You go in fresh.
- Reads your watch history file (every title you've rated), your taste profile (what earns 8+, what lands below 5, and why), and your hard exclusion rules
- For TV shows, checks the entire series run — not just the pilot — before recommending
- Never recommends anything already in your history or watchlist
- Output is title + year only, no genre hints or descriptions — minimizes expectations bias
- Companion commands:
/rate,/watchlist,/similar,/exclude,/taste,/stats,/wildcard
Defines a falsifiable hypothesis before anything runs. Asks questions in rounds until the experiment is fully specified. Builds a compute budget with itemized overhead. Writes the plan to a handoff doc. Then stops.
- Asks at least 3–4 rounds of questions before proposing anything — "something will improve" is not a hypothesis
- Forces five fields: hypothesis statement, independent variable, dependent variable, control, and a concrete success criterion (a threshold, not "better")
- Estimates compute budget: total inferences × time per inference × cost, itemized — no flat percentage buffers
- Designs sample selection (stratified, with positive AND negative samples), variants (always includes a control), and iteration strategy
- Presents the complete plan for explicit approval before writing anything to disk
Reads the approved plan. Asks questions about the actual codebase before writing code. Writes a script with dry-run mode, checkpointing, and ETA logging. Gets explicit approval before executing. Always runs in the background.
- Asks about file paths, existing code to reuse, data format, and edge cases before writing a single line
- Every script must include:
--dry-runflag, checkpoint saves every N inferences, progress logs with ETA, and resume-from-checkpoint logic - Single inference failures never crash the run — caught, logged as null, execution continues
- Dry-run first: you see exactly what will execute and how many inferences before any compute is spent
- Asks "go ahead?" explicitly — never executes without a yes. Runs in background once approved.
Reads the results. Computes metrics for every variant. Shows per-sample breakdowns. Checks against the success criterion. Discusses findings with you. Recommends ADOPT / VALIDATE / PIVOT / REJECT. Writes the decision doc only after you agree.
- Computes confusion matrix (TP/FP/FN/TN), recall, precision, FPR, and F1 for every variant — not just the winner
- Shows absolute counts alongside percentages; flags when N < 30 and percentages are misleading
- Per-sample breakdown: which specific samples changed under which variants, not just aggregates
- Explicitly checks the success criterion: did any variant hit the threshold defined in the plan?
- Discusses before deciding — asks what stands out, what matches intuition, whether confidence is high enough to act
- Decision framework: ADOPT (criterion met, N sufficient), VALIDATE (promising but small N), PIVOT (learned something, new hypothesis), REJECT (no signal, close the line)
Interviews you about what you're building, then creates the project folder, a tailored CLAUDE.md, scaffolding, .gitignore, and git repository in one pass.
- Asks discovery questions one or two at a time: name, type, stack, whether you'll use an agent workflow, whether you want a GitHub repo
- Creates appropriate scaffolding based on project type: coding projects get
src/,tests/, andHandoffs/; content projects getdrafts/anddocs/ - Writes a CLAUDE.md tailored to the project type — the living reference document all future agents will read and update
- Initializes git and pushes to a private GitHub repo
- If you're using the dev agent workflow, the agent sequence is documented in CLAUDE.md so you always know where you are in the build
Coming Soon
The job search agents re-outfitted for active pipeline work. Company research, contact profiling, and multithreading built for AEs prospecting into target accounts — not interviewing at them.
- Company research reframed for deal intelligence: buying signals, tech stack, budget cycles, not "why work here"
- Contact profiling tuned for champion identification and multi-threading live deals, not interview prep
- Outreach angle generator produces cold sequences, not thank-you notes
- Built around the AE's workflow: account planning, QBR prep, competitive displacement, renewal risk
A Claude-powered Slack bot for your sales team. I build it, configure it, and deploy it into your workspace. Could be a daily signal digest, an AI research assistant your reps can @mention, or a custom workflow — we figure out what's actually useful for your team and ship it.
- Signal digest bot: monitors your target account list and posts funding / hiring / news alerts to a sales channel on a schedule
- Research assistant: reps @mention the bot with a company name or LinkedIn URL and get a formatted brief inline
- Custom workflow: you tell me what your team actually needs, I scope it and build it
- Done-for-you service: I handle the Slack API setup, Claude API integration, hosting, and deployment
Get the files
Each package is a folder of plain text .md files. Download it, open Claude Code in the folder, and Claude walks you through installation conversationally. No manual file moving.
Individual systems available above at listed price · Sold via Lemon Squeezy · Requires Claude Code (free)