mcpField
C
context-layer-consolidation

context-layer-consolidation

github.com/posthogProductivity95108 installsUpdated today
Repository

About context-layer-consolidation

context-layer-consolidation is a skill for maintaining a mounted organizational wiki. It guides evidence-led cleanup of sourced pages, including outdated priorities, replaced decisions, duplicates, links, and oversized pages, followed by linting.

Key features

  1. 01Lint the consolidation queue
  2. 02Remove unsupported synthesis
  3. 03Mark historical priorities and superseded decisions
  4. 04Merge near-duplicate pages with wikilinks
  5. 05Resolve disagreements only with evidence
  6. 06Run lint after editing

Use cases

  1. 01Prepare wiki pages before publishing
  2. 02Clean up outdated priorities and decisions
  3. 03Consolidate duplicate content while preserving links

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/posthog/ai-plugin --skill context-layer-consolidation

After installing, type/context-layer-consolidationin Claude Code to confirm it loaded.

Details

From the project README

Context layer consolidation

Start with scripts/lint --report. Work the queue in the mounted wiki before the dream publisher creates its branch and commit.

Only edit sourced Markdown content under org/, areas/, decisions/, and existing Space pages. Never edit repository instructions, generated indexes, or scripts/; report structural failures in the run summary for the server to repair.

  • Remove unsupported synthesis.
  • Remove expired priorities unless historically significant; mark those status: historical.
  • Mark replaced decisions superseded and link their replacement.
  • Merge near-duplicates and leave a wikilink from the old subject.
  • Resolve disagreements only with evidence; leave the rest explicit.
  • Repair or intentionally retain ghost links, add sources, and split oversized pages.
  • Name every deletion in the run summary. The dream publisher includes it in the commit message.

Keep this bounded and evidence-led. Run scripts/lint after editing.

Comments

Markdown supported