React Developer: One Dashboard Review-Panel Fix — $150 Paid Trial

PAID TRIAL: ONE EXISTING REACT REVIEW-PANEL IMPROVEMENT — $150 FIXED We are building Sam OS, a practical operating dashboard supported by AI agents. An existing React frontend, Python APIs, SQL-backed task/review records and partial integrations already exist. We want an engineer who can improve inherited code, verify actual behavior and deliver small, useful increments. We are not hiring someone to rebuild everything from scratch. THE $150 TRIAL — EXACT SCOPE Improve ONE existing LinkedIn comment-review panel to match our supplied annotated prototype: spacing, typography, source-summary hierarchy, action-button layout and a usable mobile arrangement. Not the whole dashboard, page or agent platform. We will provide the exact source baseline, annotated target and sanitized fixtures before the contract starts; agree the specific remaining discrepancies first so no completed fix is duplicated. Trial acceptance: 1. Focused source patch/branch with a short explanation of changed files. 2. Before/after screenshots at390px and1440px, no document horizontal overflow, clipped controls or unreadable text. 3. Preserve existing draft editing, Save, Copy, Approve, Skip, source links, version/conflict handling and unsaved-edit protection. Approval must never mean publication. 4. Relevant automated regression evidence with fixtures and reproducible test instructions. Existing checks remain passing; report limitations honestly. 5. No backend/schema/auth changes, production deployment, real social publishing, outreach, new paid tool or production credentials. This is a paid implementation trial, not unpaid speculative work. $150 covers ONLY the agreed panel improvement and its verification. Any expansion requires a separate scope and budget. We will review the result before deciding on further work. BROADER AGENT VISION — CONTEXT FOR POSSIBLE FOLLOW-ON WORK, NOT INCLUDED IN THE $150 We want a persistent, server-hosted workforce that turns well-defined tasks into tested deliverables with limited human intervention. It must work when the owner’s laptop is closed, recover from interruptions and ask for specific decisions instead of silently stalling. Existing capabilities are partial; a configured agent, healthy process or attractive animation is not proof of successful automation. Target responsibility split: - Owner: goals, priorities and consequential approvals. - Orchestrator/Hermes: intake, clarify requirements, check dependencies, select the right specialist and track progress. - Engineer1, Engineer2, Engineer3: implement bounded work in isolated branches/workspaces. These are intended logical roles; independent operation and safe parallel capacity must be verified, not assumed. - Reviewer/Guardian: evaluate requirements, tests, evidence and safety separately from the author. - Research, writing and other specialists: scoped supporting tasks where they add value; avoid redundant always-running agents. Target automation levels (our project stages, not an industry certification): A. Assisted work: agents research, draft or propose changes; human decides what proceeds. B. Bounded execution: approved tasks run inside defined tools, workspace, usage and time limits; durable results return for review. C. Supervised autonomous queue: eligible Ready tasks are claimed automatically by the right available worker, built, tested and independently reviewed. Human-owned decisions and blocked work are excluded. D. Controlled improvement: agents identify recurring bottlenecks, research maintained repos/skills and propose measurable improvements. They do not self-authorize installations, broader access, spending or deployments. Task and persistence requirements for later increments: - Kanban is the execution record. Tasks need project/category labels, deliverable, owner, acceptance criteria, inputs, dependencies and approval boundaries. - Retain installed native statuses. Testing can be an evidence phase; do not invent incompatible scheduler states. Ready-to-running-to-review must produce actual artifacts and receipts. Blocked work needs a reason, responsible owner and next action. - Requirements workbook is a linked goals/decisions/capability index, not a competing task database. - Persist task/run identity, checkpoints, artifacts, test results, consumed usage, approvals and review outcomes outside model conversation memory. - A context reset, gateway restart or closed browser must not erase ownership or require reconstructing the project from chat. Resume from concise persisted checkpoints. - Prevent duplicate claims and external effects; use explicit run identity, ownership/leases, idempotency and bounded retries where appropriate. Unknown external outcomes hold for reconciliation rather than blindly retrying. - Human approval is scoped to the actual action/version. It is not a universal future permission; approval and execution are distinct. Review completion does not automatically authorize deployment. - Expose truthful last-run times, failure/hold reasons and meaningful notifications. Distinguish configured, reachable, running and successfully completed. Unknown costs or missing data must remain Unknown, not0. - Keep worker credentials/access limited, repositories and records owner-controlled, and tool/provider costs bounded. Paid model fallbacks stay off unless explicitly authorized. Publishing, outreach and production deployments remain gated. DELIVERY APPROACH Reuse existing/native capabilities and maintained components before custom code. Review license, compatibility and operating cost. Work through existing Kanban items, avoid duplicate systems, ship a small complete slice, test it and document rollback. AI coding assistance is welcome, but you own diagnosis, code quality and verification. Broader Python/API/Linux/background-worker experience is valuable for later work; it does not enlarge the trial. PLEASE RESPOND WITH - One relevant React interface you personally improved and what you changed. - How you would preserve existing behavior while changing presentation. - A brief example of a reliable background/agent workflow you delivered, if applicable, distinguishing your contribution and what was actually deployed. No unpaid design or code exercise requested. - Your turnaround, availability and confirmation that the narrow trial fits $150 fixed. Potential follow-on work will be separately prioritized and agreed after the trial. No ongoing volume, full-time role or additional budget is promised.