Skip to main content

JobJitsu Features & Modules

Module map for the AI Career Operating System. Describes what exists in the product vision, not how to build it.

Parent: PRODUCT_VISION.md · Sequence: ROADMAP.md · Guardrails: NON_GOALS.md · PRINCIPLES.md · TERMINOLOGY.md

Brand note: metaphor belongs in microcopy; navigation uses plain nouns (Applications, Follow-ups, Agent) — see ../brand/DESIGN_SYSTEM.md.


Feature status​

StatusMeaningContributor expectation
CoreSpine of the Career OS / v1 pathPrefer PRs here; regressions block release
ExperimentalSpecified or in-tree; API/UX may change; may be flaggedWelcome; no stable-API promise
FutureVision-accepted; not near-term implementationSpec OK; implementation deferred unless status is promoted

Horizons (ROADMAP.md) answer when. Status answers commitment. Promote Future → Experimental → Core only after the module admission test.

Functional detail: PLATFORM_SPECIFICATION.md. Shell IA plan: SHELL_IA.md.


Core modules​

The spine of the Career OS. One product, one voice, one privacy model. Together they fulfill the Leverage Belt pipeline: search → curate → apply → send → follow up.

ModulePurposeBrand postureStatus
Identity & ResumeOn-device profile, Paths, and Resume Library for tailoringStored locally; never “synced to JobJitsu cloud”Core · H1
PreferencesRules for fit, tone, constraints, approval gates“Your rules, your signal”Core · H1
Local IntelligenceAI Provider + Context Builder for roles and draftsStatus chrome: Agent · On-deviceCore · H1
AgentPrep work against preferences; pause/resumeServes; does not own send. Idle: belt tiedCore · H1
Discovery & CurationFind and filter roles via Job ProvidersPrecision first, volume secondCore · H1
ApplicationsDraft, tailor, version, track each throwCraft unit of the OSCore · H1
Queue & ReviewHolding area before anything leavesApproval is a first-class ritualCore · H1
SendExplicit outbound apply / submitUser-initiated; quiet successCore · H1
Follow-upsPolite nudges on a calm calendarAmber caution, never shameCore · H1
Timeline & MemoryLocal history of actions and outcomesLearning without leaderboard energyCore · H1
Privacy & Trust ChromeBadge, logs, “what left / what stayed”Architecture made visibleCore · H1

Core module intents (vision-level)​

Identity & Resume — Materials live in the dojo. Tailoring reads from the Resume Library; the user remains author of the final voice.

Preferences — Fit rules and approval gates are first-class. The OS does not invent urgency to override them.

Local Intelligence — On-device AI Provider by default. Status chrome shows Agent · On-device; “Local LLM” belongs in advanced settings and technical docs.

Agent — Prepares applications and reminders under preferences. Never owns the send button.

Discovery & Curation — Surface roles worth a throw via Job Providers; filter toward fit, not feed addiction.

Applications — Each application is a unit of craft: draft, tailor, track (see Application Pipeline in the platform specification).

Queue & Review — Nothing high-stakes leaves without review when the user requires it. Distinct from the AI Task Queue.

Send — The moment data may leave the machine — explicit, confirmed, calm.

Follow-ups — Gentle nudges timed to the hunt, not chase-the-recruiter aggression.

Timeline & Memory — Continuity of craft: what was sent, nudged, learned.

Privacy & Trust Chrome — Always-on assurance that the OS is still local and inspectable.


Experimental​

In-tree or specified capabilities that may change. Label clearly in UI and docs. Architecture contracts: AI_ARCHITECTURE.md (workflow engine section), DATA_MODELS.md.

Module / capabilityWhy experimentalGuardrail
Knowledge BaseGrounded retrieval for Context Builder; backlog epic pendingDistinct from Timeline; local only
AI ValidationPost-generate gates; backlog epic pendingFailures never enqueue Send
Workflow / Task QueueDeclarative AI control plane; backlog epic pendingNever call Send; ≠ review Queue
Browser automation (apply assist)High variance across sites; approval UX still evolvingNever bypass Queue → Send; user approval required by default
User-configured remote AI ProviderOpt-in cloud model endpointHonest chrome — never labeled Agent · On-device
Trusted AutomationReduced per-action approval when user enablesDefault off; no silent irreversible egress; Timeline always records Send.*; cannot weaken Send honesty
AI PlaygroundPrompt/provider experimentationMust not silently affect production Workflows
Email / Gmail deep syncPE20 in progress (@jobjitsu/mailbox) — OAuth loopback in the desktop appOpt-in OAuth; local classify/match; never auto-send. Do not duplicate in send. Guide: GMAIL_AND_OUTLOOK.md.

Future modules​

Candidates for Horizons 2–4. Each must pass PRINCIPLES.md and NON_GOALS.md before promotion.

ModuleWhy it belongsGuardrailStatus
Outcomes & reflectionTurn applications into craft feedbackNo guilt for quiet weeksFuture
Interview readinessExtend technique past the sendPractice aid — not “guaranteed offer” theaterFuture
Narrative studioKeep story consistent across resume, letters, talksOn-device drafts; user owns final voiceFuture
Network nudgesGentle relationship follow-throughConsent-heavy; never spam-your-contactsFuture
Role-fit compassReflect skills/values against rolesGuidance, not destiny assignmentFuture
Offer & decision journalCompare offers calmlyNo high-pressure countdown UXFuture
Skills & learning mapLink gaps to intentional growthOptional; not a course marketplaceFuture
Extensible agent skills (Plugins)Open-source community techniquesInspectable; user-enabled onlyFuture
Host Extensions ecosystemDiscovery/send/UI contributionsCapability-gatedFuture
Multi-profile / multi-pathHard multi-identity for career pivots (day-to-day role faces = Path under PE03)Still local-firstFuture
Export & portabilityLeave with your data anytimeReinforces sovereigntyFuture

Future modules arrive as calm rooms in the same dojo, not as a suite of shouty micro-products.


Module admission test​

A module joins Core, Experimental, or Future only if:

  1. It keeps data on-device by default (or redesigns until it does).
  2. It increases leverage, not only volume.
  3. The user retains control of outbound actions.
  4. It reduces anxiety rather than manufacturing it.
  5. It deepens the Career OS rather than a disconnected side quest.
  6. It does not violate NON_GOALS.md.