okf

OKF knowledge

Read the applicable repository instructions first. They remain the authority for when to consult memory and how to resolve conflicting guidance.

1. Identify the active bundle and scope. With the optional CLI, run irudd-okf context. Otherwise follow the bundle path in the repository instructions. Do not search unrelated repositories or infer personal bundles.
2. Read its root index.md if useful. Search for the current task, relevant code area and likely failure cases. Retrieve a few source files; avoid loading all rule titles, all skill triggers or the full corpus.
3. Follow relevant links and inspect the actual code or documentation cited by a rule. A search result is a candidate, not evidence that every relevant rule has been found. Refine the query when names or task scope change.
4. Apply the retrieved guidance under the repository's own conventions. Keep the bundle name and file path in any explanation of a rule. If conflicting rules affect the task and the instructions do not resolve them, explain the conflict and ask the operator. Do not invent a priority hierarchy.
5. When asked to add a rule, or when repository instructions authorize learning, record one specific, reusable lesson with enough context to act on it. Avoid duplicates by searching first. Personal edits require explicit operator authorization; repository edits remain ordinary files in the code review diff.

CLI discovery follows a short sequence:

irudd-okf cli search "find rules"
irudd-okf cli schema search
irudd-okf search "migration tests" --scope repo --limit 5
irudd-okf read repo architecture/migrations.md

Without the CLI, use ordinary file search (rg) and read the relevant Markdown. For safe authoring, format details, and conflict examples, see references/usage.md. Native skills remain useful for task procedures; OKF stores the facts and rules those procedures may retrieve.