Skip to content
Uvilo OS
published Visibility internal Owner erik@uvilo.com Approver _ Created _ Updated _

Project Extract

When to Use

Project Verify is complete and knowledge needs to be extracted before completing the project.

Inputs

  • Project path: {Dept}/Projects/{Project}/

Context Boundary

Read only: Vision, Spec, Changelog, and work products (implemented files, reference files). Never read Research or Plans — those are historical.

Prerequisite

Must be a fresh agent session — new conversation with clean context.

Procedure

  1. Read: Vision, Spec
  2. Update {Project}_Phase.md: set Phase: Extract_Started
  3. Build, commit, push
  4. Identify extractable content
    • System documentation → department root-level Knowledge files like Forge/Forge_Infrastructure.md or Product/The_Uvilo_Method.md
    • Reusable procedures → department Skills/ folder
    • Use the Changes table in Changelog to identify all work products that may contain extractable knowledge
    • For each candidate, verify it is not already present in a permanent doc. If the only copy lives inside the project folder, it’s extractable. If permanent docs already cover it, skip it.
  5. Decide best way to extract content
    • Sometimes the Spec (with edits) becomes published Knowledge
    • Sometimes changes are folded into existing Knowledge
  6. Remove duplication
    • Multiple Knowledge files must not describe the same thing
    • Secondary sources should refer to the canonical Knowledge file & heading
  7. Ensure everything is findable
    • System-wide skills must be listed in the Persistent Knowledge prompt (Bot database)
    • Department-level knowledge should be linked to from {Dept}/AGENTS.md and should have status published
  8. Build, commit, push
  9. Proceed directly to Extract_Eval — do not wait for user approval before evaluation
  10. After Extract_Eval converges, present the extraction to the user for confirmation if any user-facing decisions remain
  11. For each extraction, spawn the Page Manager bot via spawnAgent with { botGroup: 'forge', botHandle: 'page-manager' } to create new documents or update an existing document
  12. Verify: no useful information remains trapped inside the project folder
  13. Update {Project}_Phase.md: set Phase: Extract_Completed
  14. Build, commit, push

Post-Extract

After extraction is complete, the project proceeds to Extract_Eval for review of the extraction quality.

Rules

  • Every piece of useful information must reach a permanent home
  • Don’t extract trivially obvious content — focus on hard-won knowledge
  • Content already correctly handled by code is not extractable. Extraction is for knowledge that would be lost if the project folder were archived — not for bugs already fixed, decisions already reflected in code, or status already absorbed into permanent docs.
  • Extracted content must be self-contained in its new location (no dangling references to the project)
  • Context boundary is defined above
  • Do not update the status of project level files at this phase - that’s for Project_Complete