Claude-Career-Plugin

Sweep all (or a filtered subset of) email addresses in the workspace CRM and re-verify them via Hunter Email-Verifier. Use periodically — people change jobs, addresses go stale, accept-all status changes. Updates the `verified` and `last-verified` columns in `crm/contacts.md` and prints a summary of what flipped.

Scan the user's existing Claude Code plugins and MCP servers to detect capabilities the career plugin needs (email send, transcription, calendar/tasks, semantic store, CRM). For each capability, decide whether to reuse an already-installed plugin/MCP, install a recommended companion plugin, or add an MCP server. Writes the resolution map to ${CAREER_DATA_DIR}/config.json under a `capabilities` block. Invoked from /career:onboard before install-companion-plugins so we don't double up on capabilities the user already has.

Enrich an existing CRM contact with Hunter Combined-Enrichment — pulls role, seniority, department, social handles, and the contact's company facts in one call. Use to flesh out a thinly-populated CRM row before drafting outreach or prepping for a meeting. Operates on an email address from the CRM; updates the row in place.

Given a person's name and their company (name or domain), return Hunter's best guess at their email address with a confidence score. Use when the user already knows who they want to reach and just needs the address. Differs from find-contact (which lists candidates at a company) — this is targeted at a single named person.

Validate a single email address via Hunter Email-Verifier. Use when the user has an address (handed to them, scraped, guessed) and wants a deliverability + confidence read before sending outreach. Returns Hunter's verdict (deliverable / risky / undeliverable / unknown), score, and the disposable / role / accept-all flags. Does not write to the CRM unless asked.

For CONTEXT_STORE=pinecone — verify the index against the local workspace. Flags vectors whose source file was deleted, renamed, or content-changed (hash mismatch), and surfaces vectors whose content contradicts current ground-truth (stale claims, closed engagements, outdated targets). Proposes deletes/re-embeds; never silently mutates the index.

First-run scaffold of a vector-store namespace dedicated to application/outreach tracking. Creates the `applications` namespace in the configured index, sets the metadata schema (company, role, status, applied_date, last_touch, source_path, stage), seeds it from crm/outreach.md + crm/opportunities.md, and registers it in config. Idempotent — re-running detects existing namespace and offers reseed or skip.

Flat/local-mode workspace hygiene. Sort loose files into canonical folders (research/, domain-notes/, inbox/, crm/, resume/, snippets/, branding/, cache/), find near-duplicates by name+hash, flag stale inbox entries (>14d), and re-root mis-placed files. Proposes a plan, applies only on confirm. No-op when CONTEXT_STORE != markdown.

Produce a Company Financials research brief — funding rounds, valuation, revenue (if disclosed), burn / runway, profitability signal, public-market metrics. Use when the user needs a numbers-only read for negotiation, due-diligence, or to gauge runway risk. Reads templates/research-briefs/company-financials.md and writes to <WORKING_FOLDER>/research/companies/<slug>/company-financials.md. Numbers are sourced or marked "not disclosed" — never estimated.

Produce a Company Overview research brief for a target company — name, locations, headcount, funding rounds at a glance, founder story, competitive landscape, vision, differentiators. Use when the user wants a top-level read on a potential employer or consulting client. Reads the template at templates/research-briefs/company-overview.md and writes to <WORKING_FOLDER>/research/companies/<slug>/company-overview.md. Cites at least 3 sources. Skipped fields are marked unknown, never fabricated.

Produce a Cultural Fit research brief — pairs the user's stated operating style and constraints (read from ground-truth.md) against the company's apparent operating mode (from public signals). Output is a match-map plus blunt friction list. Use after the other briefs are in hand. Requires populated ground-truth.md. Reads templates/research-briefs/cultural-fit.md and writes to <WORKING_FOLDER>/research/companies/<slug>/cultural-fit.md.

Produce a Glassdoor / employee-sentiment research brief — overall ratings, sub-ratings, tenure data with red-flag detection scoped to the relevant team, compensation data, theme analysis of top complaints and positives, recent direction inflections. Aggregate sentiment, not a recap of individual reviews. Use to gauge whether public employee discourse aligns or contradicts the company's outward image. Reads templates/research-briefs/glassdoor-signal.md and writes to <WORKING_FOLDER>/research/companies/<slug>/glassdoor-signal.md.

Produce a Recruitment Profile research brief — careers page, open roles (filtered by target function if provided), LinkedIn presence, headcount trend, hiring signals, recruiter activity, and known interview-process patterns from public sources. Use when the user wants to assess whether a company is actually hiring people like them, vs. posting ghost roles. Reads templates/research-briefs/recruitment-profile.md and writes to <WORKING_FOLDER>/research/companies/<slug>/recruitment-profile.md.

Produce an honest Remote Friendliness research brief — does the company actually hire from the user's location, or is the "remote-first" claim marketing? Cross-references stated policy against live job postings, time-zone requirements, hiring entities, and existing employees in the user's region. Critical for non-US/EU users. Reads templates/research-briefs/remote-friendliness.md and writes to <WORKING_FOLDER>/research/companies/<slug>/remote-friendliness.md.

Given one seed company, find peers along an explicit similarity axis (product, customer, tech stack, business model, culture). Writes to <WORKING_FOLDER>/domain-notes/like-<seed-slug>.md. Reads outreach.md to mark contact status. Use when the user has a known reference point and wants more of the same shape.

Enumerate the main companies in a defined niche (e.g. "agentic AI for policy simulation", "open-source LLM tooling"). Writes to <WORKING_FOLDER>/domain-notes/<domain-slug>.md. Reads outreach.md to flag already-contacted companies. Use as the first step before /career:suggest companies — populate the domain knowledge once, suggest from it many times.

Enumerate companies in a domain that are genuinely remote-friendly to USER_LOCATION (not just "remote-OK" with US-only fine print). Writes to <WORKING_FOLDER>/domain-notes/remote-<domain-slug>.md. Honest negatives over hopeful positives.

Generate an outreach email draft grounded in ground-truth.md, the company research brief, the target contact, and a chosen template (cold-pitch / warm-followup / consulting-pitch / job-followup / interview-followup). Always saves to drafts/ and presents to the user for review. Never sends. send-outreach is a separate skill that requires explicit confirmation.

Find the right person to contact at a target company using Hunter.io. Use when the user needs decision-maker / hiring-manager / founder contact details before drafting an outreach. Returns ranked candidates with confidence scores; optionally appends to the workspace CRM. Hunter MCP is required.

Read or edit the workspace's ground-truth.md — the source-of-truth profile every Career-OS skill reads first. Use when the user says "update my profile", "I changed roles", "I'm looking for X now", or after a major life/career change. Idempotent and diff-friendly. Never silently overwrites.

Offer to install Claude Code plugins that complement the career plugin. Use during /career:onboard, or whenever the user wants to enable a downstream capability (transcription for voice ingest, schedule-manager for calendar/tasks, email-skills for outreach send, social-feedback for company sentiment research, decision-evaluation-framework for offer comparisons, private-misc for Pinecone wiring). Lists each, explains what it unlocks in the career plugin, and lets the user accept/decline per plugin.

Manually log an outreach that was sent outside the plugin (e.g. a LinkedIn message, an in-person conversation, an email sent from a phone). Appends a row to crm/outreach.md without invoking any send. Use this to keep the outreach log honest as the spine of the recommendation engine.

Send a previously-drafted outreach email. Reads a draft file from drafts/, presents it one final time to the user, requires explicit "yes, send" confirmation, then delegates to the email-skills companion plugin (personal or business sender per draft frontmatter). On send, updates the draft frontmatter to status=sent and appends a row to crm/outreach.md. Never sends without explicit confirmation.

Update the status of an existing outreach log row. Use when a contact replies, books a meeting, the meeting happens, the conversation stalls, or the opportunity closes. Records state-change date without rewriting history. The recommendation engine reads these statuses to enforce anti-loop discovery.

Accept an audio file path, transcribe via the claude-transcription companion, classify the transcript (quick-note / retrospective / strategic-dump / unclear), and propose a route. Never auto-routes — always confirms with user. Lands transcript + audio link in <WORKING_FOLDER>/inbox/<YYYY-MM-DD-HHMM>.md.

Audit the user's public brand surfaces — personal site, GitHub, Hugging Face, LinkedIn — against ground-truth direction. Surfaces drift (site says X, ground-truth says Y) and gaps (no project shipped in domain Z). Produces an actionable update list.

Two-way sync between markdown CRM (crm/companies.md, crm/opportunities.md, crm/outreach.md) and Airtable. Reads CRM_ADAPTER=airtable. Markdown stays the source of truth — Airtable is a mirror with a diff back. Conflicts surfaced for user resolution, never auto-merged.

Two-way sync between markdown CRM and Attio. Same shape as crm-adapter-airtable but mapped to Attio's object model (Companies, People, Deals, Notes). Markdown stays canonical; Attio is a working surface. Conflicts surfaced, never auto-merged.

One-shot migration between CRM backends. Markdown → Airtable, markdown → Attio, Airtable → markdown, Attio → markdown, or Airtable ↔ Attio. Use when first adopting an external CRM, when leaving one, or when swapping providers. Snapshot-style, not incremental.

Incremental sync of working-folder/ into a Pinecone index for semantic recall. Embeds only changed files (mtime + content hash). Index name from config. Delegates first-time MCP setup to private-misc:personal-context-injection. Idempotent.

Cost-routing helper. Reads RESEARCH_MODEL from config and decides whether a bulk-research call should run on Claude or be delegated to a cheaper model (DeepSeek Pro/Flash via OpenRouter). Reserves Claude for ground-truth, meeting-prep, draft-outreach, recommendation engine. Skills that produce many cheap research artifacts (discovery, ecosystem, financials briefs) call this first.

White-hat salary benchmarking for a (geography, role) pair. Cached at cache/salary-<geo>-<role>.md so repeat calls don't re-fetch. Public sources only — postings, Levels.fyi, Glassdoor, Payscale, BLS/Eurostat, SEC, Blind/Fishbowl as signal. Confidence-tagged per data point.

Query Pinecone (or fallback to grep-over-markdown) for context relevant to a task. Returns ranked snippets with cited paths. Used by suggest-companies, meeting-prep, weekly-plan, self-healing-context when CONTEXT_STORE=pinecone. Caller decides what to do with results.

Diff current skill inventory against target-role/domain profile. Categorises gaps as critical/material/met. Carries plan notes forward across refreshes. Writes skills/gaps.md. Different shape from suggest-skills (which is forward-looking against a horizon) — this is the inventory delta.

Produce a resume variant tailored to a specific role/JD. Reads master resume + ground-truth + JD; rewrites bullets to surface relevant experience and mirror JD phrasing where truthful. Never invents experience. Writes to resume/variants/<company>-<role>.md with a tailoring summary.

Lightweight capture for inline text or short voice notes — same classification + routing pipeline as voice-ingest, but optimized for "one thought, drop it in" rather than full memos. Voice routes through the transcription companion; text skips straight to classify + propose route.

Reconciliation pass. Reads ground-truth.md + crm/companies.md + crm/opportunities.md + crm/outreach.md and surfaces discrepancies. Never silently overwrites — every finding is a proposed delta the user reviews. Writes to <WORKING_FOLDER>/reconcile/<YYYY-MM-DD>-findings.md.

End-of-week wrap-up. Reads outreach diffs since last log, reads the week's plan, asks user what worked, what didn't, what's next. Produces a wrap-up doc. Delegates to schedule-manager wrapup-style if available. Saves to <WORKING_FOLDER>/plans/<YYYY-MM-DD>-week-log.md.

Read crm/outreach.md + crm/opportunities.md + recent recommendations, propose 5–7 actions for the week (mix of research, outreach, follow-up, meetings, ecosystem). Delegates actionable items to schedule-manager:batch-create-tasks and time-blocked work to schedule-manager:batch-create-events. Confirms before any external write.

Enumerate communities (LinkedIn groups, Slack channels, Discords, professional orgs, mailing lists) for a domain. Each entry includes activity-level signal and alignment to ground-truth. Writes to <WORKING_FOLDER>/domain-notes/<domain-slug>/communities.md.