Reference
Skills Catalog
All 28 skills and what each one does.
CareerRat ships 28 skills under .agents/skills/. Each is a SKILL.md file, a
detailed, executable procedure the agent follows when routed to that skill. Skills
are field-neutral: the same skill serves any candidate because it reads your
private config, not hardcoded preferences.
Apply cycle
| Skill | Description |
|---|---|
ingest-profile | Interview a new candidate to produce all user-layer config files: profile, targeting, evidence, honesty boundaries, form defaults, writing style, personalized AGENTS.md, and seed search sources. Run on first setup or any major profile change. |
setup-searches | Build and maintain config/search-sources.yml from candidate targeting — generate a baseline from role buckets, curate enabled sources, add or import individual searches (including pasted board URLs with embedded filters preserved), and tune global filters. Upstream of search-jobs; does not scan. |
search-jobs | Run the top-of-funnel sourced sweep — scan configured sources, dedupe, check liveness, coarse-triage every new entry with fitScore/fitBucket/fitBasis, save JD bodies, write watermarks, and optionally hand top sourced roles to evaluate-job. Does not tailor, fill, or submit. |
evaluate-job | Body-read gate for a single job posting — fetch, save, check limits/comp/fit, emit GATE/FIT/COMP/ACTION, and write calibrated fit back to the tracker. Run before any tailoring or submission. |
answer-question | Draft evidence-grounded answers to one-off application or screening questions, reuse saved disclosure answers, and return NEEDS YOU instead of guessing when the evidence is insufficient. |
tailor-application | Generate finished, honest role-specific resumes, cover letters, short answers, and non-message outreach artifacts from a saved JD, source resume facts, candidate evidence bank, writing style, and form defaults after the job passes the CareerRat gate. Reject unresolved placeholders. |
apply-job | End-to-end job application workflow: gates on app limits and evaluate-job verdict, tailors artifacts, builds the render, prepares supported ATS forms (LinkedIn jobs go to the employer's own link, or wait for you), submits when complete and verified (or stops for you after careerrat automation submit manual --write), verifies the outcome, and updates the tracker. |
coach-gaps | Turn each named REVIEW fit gap into an evidence-grounded claim draft or an explicit no-close path, then offer re-evaluation only after the evidence firewall confirms the change. Runs on its own when a review verdict lands. |
Intake extraction
| Skill | Description |
|---|---|
intake-extract | Backend-only verbatim transcription for PDF and image uploads sent through Universal Intake. It does not summarize, classify, or make candidate-specific assumptions. |
resume-extract | Backend-only structured extraction for PDF, image, and converted DOCX résumés. It emits the onboarding profile and targeting draft without inventing facts. |
Research loop
| Skill | Description |
|---|---|
research-company | Web-search a company across six domain-neutral axes and compose a cited .careerrat/workspace/research/<slug>.md artifact that feeds interview-prep and evaluate-job FIT context. |
research-comp | Web-search market comp for a role+location and produce a cited .careerrat/workspace/research/comp-bench-*.md that feeds evaluate-job COMP CHECK/ANCHOR and a confirm-first gate write-back. |
research-boards | Web-search and legitimacy-screen new job boards for the candidate's domain/role families, then propose adding to search-sources.yml confirm-first. |
discover-companies | Web-search companies likely to be hiring the candidate's target roles, resolve each to a scannable ATS careers board, legitimacy-screen and dedup, then propose adding to config/sourced-scan.json tracked_companies confirm-first. Upstream of search-jobs. |
company-health | Score a role-scoped company-risk rating (healthy|watch|risky) from web-searched layoff, hiring-momentum, financial, sentiment, and leadership signals with provenance; persist it to the tracker and feed the fit score only where it cross-cuts a stated candidate need. Runs on its own at the fire stage; an internal signal that never enters an outbound artifact. |
Communications
| Skill | Description |
|---|---|
email-comms | Draft, reply to, follow up on, summarize, and track job-search email and recruiter-message threads — including thank-yous, scheduling, negotiation, and cold outreach — while persisting full thread context in the CareerRat tracker. |
schedule-meeting | Handle recruiter / hiring-team scheduling threads — read the proposed slots, the candidate's timezone and availability, and any calendar context; draft a clear availability reply, prepare a calendar-ready hold, and write scheduling state back to the tracker. Confirm-first before any outbound reply or calendar write. |
ingest-mail | CareerRat's in-app workflow reads job-search updates from Apple Mail on macOS or Gmail/Outlook webmail while mail_access is on; it captures relevant threads and source watermarks locally. The terminal skill provides the same workflow for external agents. Read-only at the source. |
ingest-messages | CareerRat's in-app workflow reads LinkedIn and Wellfound DMs through the session browser and captures relevant threads and source watermarks locally. The terminal skill provides the same workflow for external agents. User-initiated and read-only at the source. |
Interview & outcomes
| Skill | Description |
|---|---|
interview-prep | Build interview packets, script live comp answers, and capture post-interview debriefs and durable lessons from a saved JD, tracker record, candidate evidence, and interview invite context. |
track-outcomes | Record application outcomes, execute status transitions in tracker.json, append durable patterns to role-family learning files, check reevaluation thresholds, and write back any mid-flow gate changes stated by the user. |
reevaluate-strategy | Learn from accumulated outcomes — read the funnel, rejections, interview transcripts, and wins, then recommend concrete tuning to targeting signals, comp anchoring, fit calibration, channel mix, and writing-style. Triggered by outcome thresholds or an explicit review request. |
Optimization & sync
| Skill | Description |
|---|---|
optimize-linkedin | CareerRat's in-app workflow reads the candidate's LinkedIn profile and proposes honest, evidence-backed rewrites for review. Approval stays local; the terminal skill writes approved edits live only with per-field confirmation and the profile_apply capability on. |
sync-status | CareerRat's in-app workflow reads saved ATS dashboards, normalizes each raw label, applies verified high-confidence advances atomically, and leaves regressions or low-confidence labels for review. Terminal agents route confirmed transitions through track-outcomes. User-initiated and read-only at the portal. |
relationship-sourcing | CareerRat's in-app workflow finds likely recruiters, hiring-team members, or warm contacts through LinkedIn and Wellfound, then saves review-only leads in Network. The terminal skill provides the same workflow for external agents. User-initiated and local-only. |
Settings & calendar
| Skill | Description |
|---|---|
configure | Show current settings and route changes to the validated CLI or owning skill — never a new mutation path itself. |
calendar-sync | Write tracker-derived interviews, assessments, follow-ups, prep blocks, and deadlines to Apple Calendar, Google Calendar, Outlook, or approved local automation tools. User-initiated, confirm-first, and off if you turn calendar_sync off. |
Diagnostics
| Skill | Description |
|---|---|
report-issue | When CareerRat itself looks broken — a crash, stack trace, non-zero exit, or clearly-wrong output — separate a real defect from a config problem, assemble redacted diagnostics (no candidate PII, comp, or workspace contents), and, only with an explicit yes, open a GitHub issue on the upstream CodesWhat/careerrat repo under the user's own gh identity. Never auto-files; falls back to a prefilled issue URL when gh is unavailable. |