e0ipso/kenkeep/bootstrap

/kk-bootstrap skill

Supervised, agent-driven, one-pass bootstrap. The skill body lives at src/templates-source/skills/kk-bootstrap/SKILL.md (the bytes are identical across all five harnesses; the SKILL.md resolves the active harness at runtime via the /tmp/kk-detect-harness.mjs helper).

/kk-bootstrap                      # scans docs/ and root *.md
/kk-bootstrap docs/architecture    # scope to a path

The skill surveys markdown, splits content into practice and map nodes, and writes them directly under .ai/kenkeep/nodes/. It is judgmental, not exhaustive — samples, follows cross-references, stops to ask the user when scope is unclear. Existing nodes are never overwritten; collisions are skipped and reported.

Unlike bootstrap-incremental, the skill runs through the Task tool (a sub-agent), not a headless subprocess. Consequence: it honors bootstrapModel.name on a best-effort basis but ignores bootstrapModel.effort because the Task tool has no effort parameter.

Reviewer workflow: git diff nodes/, accept individual files with git add nodes/<folder>/<file>.md && git commit, reject the rest with git restore nodes/<folder>/<file>.md.

For re-runs after editing docs, use the headless bootstrap-incremental CLI instead (hash-aware, deterministic chunking).

<!-- kk:related:start -->

Related

• Related: map-bootstrap-incremental-command
• Related: practice-bootstrap-never-overwrites-existing-nodes
• Related: practice-bootstrap-is-supervised-and-judgmental

<!-- kk:related:end -->

<!-- kk:citations:start -->

Citations

[1] docs/installation.md [2] docs/daily-use.md <!-- kk:citations:end -->