linkedin-featured-proof

Plan or create LinkedIn Featured case studies, PDFs, and PNG proof cards from verified work; not profile rewrites or feed content.

SKILL.md

GitHub
<!-- GENERATED BY generate_skill_wrappers.py; DO NOT EDIT MANUALLY. -->

Generated Claude Code skill — do not edit manually

This complete skill and its bundled resources are generated from agent-plugins/skills/linkedin-featured-proof/SKILL.md. Manual changes in this directory will be overwritten the next time the generator runs. Edit the canonical skill instead.

LinkedIn Featured Proof

Turn verified work into inspectable public proof for a specific LinkedIn audience. Create only the selected assets; an asset request does not require rebuilding the profile or starting a content calendar.

Establish the brief

Read current project instructions, relevant available memory, candidate-source configuration, and personal voice rules. Memory locates evidence; it cannot replace current sources or authorize publication. Do not write memory unless requested. Use proof brief and provenance for evidence rules, job-intelligence integration, and the asset acceptance checklist.

Determine the audience, intended next action, strongest work evidence, public-use boundaries, language, existing brand, desired formats, and output destination from the request and supplied context. Ask one bundled question for essentials that cannot be inferred, while continuing independent preparation. Do not request a full CV or LinkedIn export for a single project card if project evidence suffices.

Choose the operation:

  • Plan: asset selection, outline, evidence gaps, and order; no file generation.
  • Copy: finished document/card text and captions; return in chat by default.
  • Produce: requested editable source plus PDF/PNG outputs where tools permit.

Select proof and write the asset

  1. Choose up to three complementary Featured items when useful: a concise expertise overview, one substantial case or work sample, and a next step aligned with job search, clients, or networking. Do not force three or five assets from thin proof.
  2. Retain an internal claim ledger with final text, exact source quote, source locator/revision, verification status, employer/role/period, contribution scope, qualifications, and public-use status. Check the source's meaning. Only verified claims with suitable public-use permission enter final assets. A source match alone does not certify an outcome or personal ownership.
  3. Make each case understandable on its own: problem and context, personal role, decision/action, result or learning, relevant technology, and inspectable proof. Use supported numbers with units and time windows; qualitative outcomes are acceptable. Distinguish team results from individual work, prototypes from production systems, and public demos from employment evidence.
  4. Use explicit role/project labels and readable text. Add a simple architecture diagram when it explains a verified decision; label simplified or illustrative elements. Never manufacture infrastructure details, scale, or client logos.
  5. Check supplied public repository, diagram, release, or portfolio destinations and contribution attribution. If a link is unavailable, label that limitation in review notes and choose another verified item or omit it. Never invent a URL.
  6. Use a goal-appropriate invitation: inspect a project, discuss relevant work, or make contact through an approved public channel. Do not expose private phone numbers, email addresses, internal URLs, or hidden metadata by default.

Produce requested files

  • Use available document/PDF/presentation capabilities and their format-specific validation when they match the requested output. Inspect existing generators before execution. Do not download or run upstream scripts, install dependencies, or upload candidate material as an incidental step.
  • Create a public export from sanitized content; keep raw source excerpts and the private claim ledger separate. Preserve existing output files and use explicit, distinguishable names for revisions.
  • Make PDFs text-selectable/searchable where supported. For PNG cards, deliver a matching plain-text caption or transcript so proof is not trapped in pixels. Use readable typography, contrast, restrained branding, and concise labels.
  • Verify current upload formats, dimensions, file limits, and the actual publishing surface before claiming upload readiness. Do not assume that a document post, Featured link, role media attachment, and Skills association behave identically.
  • Render and inspect every page/card for clipping, blank content, font problems, broken layout, unreadable diagrams, and mobile-scale legibility. Extract PDF text and check facts, selectable content, and link targets against the approved copy. Inspect export metadata and visible content for private information.
  • If rendering or export is unavailable, deliver the editable copy or source and state precisely which outputs and checks remain incomplete. A source file or an uninspected export is not a verified final PDF/PNG.

Deliver

Provide the requested files or copy, proposed Featured title/caption/order, accessible text equivalent, brief evidence/QA receipt, and any unresolved blocker. Distinguish verified facts, illustrative diagrams, and unavailable links. Keep internal evidence excerpts out of public captions. Do not claim upload success without an observed result.

Drafting and file generation do not authorize LinkedIn uploads, posts, profile edits, messages, or recommendation requests. Apply existing explicit authorization for the exact action without asking again; otherwise leave a reviewable package. Do not promise search rank, conversion uplift, interviews, or AI-parser access.