LinkedIn Profile MCP & Portfolio Evidence Capture

Problem

• Profile and portfolio updates were too easy to base on memory instead of captured proof.
• LinkedIn, portfolio, GSC, and GitLab evidence needed one repeatable workflow with explicit safety limits.
• Account-private analytics and third-party state had to remain read-only and locally controlled.

What I built

• Implemented CLI and MCP tools for profile PDF/live comparison, profile audits, browser capture, manual update packs, and portfolio/GSC evidence capture.
• Created reusable Codex skills for LinkedIn deep capture, Google Search Console capture, and portfolio evidence extraction.
• Added secret scanning, read-only guardrails, ignored local outputs, tests, and GitLab delivery hygiene.

My role

• I designed the evidence model, CLI commands, MCP tools, browser-capture boundaries, and reusable Codex skills.
• I kept owner-private LinkedIn analytics and account data out of public copy unless explicitly approved.
• I built the workflow to support profile updates without mutating LinkedIn, GSC, or GitLab state.

Constraints

• The repository and captures may include private account context, so public claims must stay bounded.
• No public adoption, external-user, or revenue claims are made.
• Private GitLab links, sessions, cookies, screenshots, and analytics remain unpublished.

Outcomes

• A repeatable evidence pipeline for making LinkedIn and portfolio updates from captured proof.
• A stronger safety model for AI-assisted profile work across browser, API, and local artifacts.
• A useful internal tool that supports SEO and AI-readable portfolio improvement without leaking private state.

Architecture decisions

• Read-only third-party automation.
• Evidence-first profile maintenance.
• Machine-readable portfolio and SEO context workflows.

Public evidence

The public record covers the read-only MCP and CLI design, reusable capture skills, local evidence packs, secret scanning, and third-party safety rules.

Proof limits

LinkedIn analytics, sessions, cookies, private screenshots, Search Console account data, GitLab tokens, and private project links are not published. Third-party mutation is outside the default contract.

Related work

AI Workflow Library · Public LinkedIn posts