linkedin-workspace

Use when setting up, migrating, synchronizing, auditing, or reporting the shared LinkedIn client-acquisition workspace across local Markdown files, Google Drive, and Google Sheets, including recording

SKILL.md

GitHub

LinkedIn workspace

Maintain the persistent operational layer for the LinkedIn client-acquisition system. Prefer one Google Drive folder plus one native Google Sheet workbook when those capabilities are available. Local Markdown files remain valid source artifacts; the Sheet is the operational index and workflow state.

Business objective

The workspace exists to measure and support this funnel:

visibility -> credibility -> conversations -> qualified leads -> paying clients

Do not optimize the system around likes, comments, or follower counts alone.

Durable workflow order

For an owned LinkedIn post, use this order:

research/evidence -> draft -> Content Library persistence -> content approval -> optional visual persistence -> ready_for_manual_post -> user publishes or schedules -> user confirms or supplies URL -> Content Library update

For comments, connection requests, and warm follow-up messages, use:

discovery/context -> exact draft -> Interaction persistence -> action approval -> live re-check -> submit -> live UI verification -> Interaction/Prospect update

The exact draft and its identifiers must exist in the Sheet before the relevant approval request. A successful LinkedIn action is never a substitute for the final verified Sheet update; if that update fails, reconcile by stable ID and do not retry the external action automatically.

Hiring-post workflows additionally use the dedicated Hiring Posts worksheet in the same workbook. Link each row to the enriched Prospect ID, preserve the exact Post URL, and link the exact response through Interaction ID. Do not create a separate workbook merely because hiring posts have a separate tab. Do not add an automated hiring-post status; Application Status is a manual field the user sets to applied or not_applied.

Public prospect enrichment is additive and provenance-led. Preserve existing user-entered values, write only professional public data or professional contact data visible in the user's authorized LinkedIn session, store field-level source and visibility evidence in Prospect Evidence, and represent missing or conflicting values explicitly instead of guessing or overwriting them. Authenticated-visible contact details are operationally sensitive: do not copy them into public comments, posts, or shared artifacts without a separate purpose and approval.

Authority hierarchy

Use these sources in this order for their respective facts:

  1. LinkedIn live UI — authoritative for whether a comment, connection, reply, or other supported interaction actually happened. This workflow does not use the UI to publish or schedule the user's own posts.
  2. Approved Markdown/source artifact — authoritative for research context, evidence, revision history, and the exact approved post body.
  3. Google Sheet — authoritative for operational state, links, prospect/interaction state, next actions, and reporting.
  4. Drive file metadata — authoritative for stored asset/file location.

For an owned post, an exact LinkedIn URL supplied by the user, or an explicit user confirmation that the post was scheduled or published, is the publication evidence this workspace may record. Do not infer publication from approval, visual readiness, or a local/Sheet row.

Never change a stronger source merely because a weaker source disagrees.

This hierarchy determines which fact wins; it does not authorize a remote mutation. Before creating a Drive folder/workbook, uploading a file, or writing Sheet state, identify the exact target, show the intended change, and confirm that the user has authorized that workspace operation. A read-only audit may inspect available state without creating or repairing remote resources.

References

Read:

Setup workflow

  1. Resolve the local workspace, if any. Never scan unrelated home directories.
  2. Detect whether an authorized Google Drive/Sheets capability is available.
  3. If an authorized write capability and user-selected target are available, create or reuse the Drive layout and the single workbook defined by the references. Do not create a separate Sheet per skill. Otherwise report the proposed layout without mutating remote state.
  4. Reuse existing tabs and columns when compatible. Add missing columns without deleting user data. Ensure the dedicated Hiring Posts worksheet exists when hiring-post work is requested; do not create a second workbook for it. Do not create or populate a separate queue tab for owned posts; preserve unknown legacy data without using it for new work.
  5. Record the workbook and root Drive folder as the workspace targets for the current run.
  6. If Google access is unavailable, continue with local/session state for read-only discovery and drafting where the requested task permits it, and clearly mark unsynchronized changes. Do not ask for approval or take an external comment/connection action from linkedin-comments or linkedin-hiring until the required draft and, for hiring work, Hiring Posts rows are persisted and verified. Do not treat a local row as proof of an external LinkedIn action.

Migration workflow

For an existing Markdown corpus, especially the user's current LinkedIn posts:

  1. Inventory filenames and metadata first. Do not rewrite content during import.
  2. Read the existing index (posts.md) when present, then parse each post file.
  3. Preserve existing stable post numbers. Convert them to stable Content ID values without renumbering. Assign IDs only where none exist.
  4. Extract the exact Final post text body as the Sheet publishing snapshot.
  5. Compute a SHA-256 hash of the exact final post text, not of unrelated notes.
  6. Re-check the authorized write capability and user-selected Drive target immediately before uploading. If either is unavailable, retain the local source and report the unsynchronized upload instead of guessing a target. Upload the original Markdown files without changing their contents.
  7. Create/update Content Library rows idempotently only after the target and write authorization are confirmed; otherwise prepare a local migration report without pretending the Sheet changed.
  8. Map old statuses conservatively. Never infer posted or scheduled from approval or visual readiness. A manually scheduled or published post becomes scheduled or posted only after the user explicitly confirms that outcome.
  9. Link research and existing visual files when they exist; otherwise leave the visual state as not_assessed.
  10. Produce a migration report with imported, skipped, duplicate, conflict, and unresolved counts.

Synchronization

When a post changes after migration:

  • increment Revision;
  • update the exact Sheet Final Post Text snapshot;
  • recompute Content Hash;
  • update the Drive source file only after re-checking the authorized capability and selected target;
  • set Sync State=conflict rather than silently overwriting when two approved revisions differ and precedence cannot be established.

For comments and connection requests, keep the existing interaction verification rules: an attempted action with an ambiguous result remains pending_verification and is never retried automatically.

For owned posts, after Approval=approved and any required visual is ready, set Content Library.Publish Status=ready_for_manual_post. The user publishes or schedules outside this workflow. When the user confirms scheduling, re-read the approved Content ID/revision/hash, set Publish Status=scheduled, record Publication Evidence=user_confirmed_scheduled, optionally record Scheduled For, leave Post URL blank, and re-read the row. When the user returns the exact live LinkedIn URL, set Post URL, Publish Status=posted, and Publication Evidence=user_supplied_url. If the URL cannot be tied to one Content ID, leave the row at ready_for_manual_post and ask for clarification. Never schedule the post or claim a live UI verification.

Reporting

Use the workbook to report useful funnel facts such as:

  • posts drafted, approved, ready for manual publication, scheduled, and posted;
  • comments verified and replies received;
  • connection requests and accepts;
  • conversations, qualified leads, calls, proposals, wins, and losses;
  • which topics/angles generated identifiable conversations or leads.

Do not fabricate attribution. If a lead cannot be tied to a specific post or interaction, record the source as unknown or mixed.