Adding in persistent knowledge base and milestone thinking - #11
Adding in persistent knowledge base and milestone thinking#11KennySmash wants to merge 2 commits into
Conversation
Add repository-native KB-Brain memory (workspaces, routing, seal/amend) and brief-to-milestone candidate-spec expansion. Existing PAAD skills are unchanged; both capabilities are invoked explicitly. Includes stdlib kb_brain.py tooling, templates, tests, docs, and regenerated exports. Co-authored-by: Kenny Webb <KennySmash@users.noreply.github.com>
…6222 feat: add KB-Brain memory and brief rumination skills
|
Hi Ovid, this is an ai assisted distillation of some battle tested techniques that I've been using over the past year in my personal projects to help combat tech debt and manage shippable milestones and thought it would be a great fit to implement something similar here, I've marked the PR as draft as its more of an idea rather than just a straight merge as I don't want to step on toes architecturally |
|
Hi @KennySmash. Sorry I didn't see this sooner. I'm on holiday for a couple of weeks, but I'll try to squeeze in investigation of this. Without digging in, my first suggestion is to fix what is probably one of my worst failures: skill names. I get them wrong all the time. For this, I would want new skill names. Many of the development teams I train are comfortable in English, but not native speakers. Thus, "kb-brain" and "brief-ruminate" are names that I can grok, but are likely to be confusing and harder for non-native English speakers to understand and/or remember. |
|
Completely fair point, I'll confer with some multi-national dev friends and get some idea's and update |
3 participants
Why
Persistent repository-native memory is needed so humans, lead agents, and sub-agents can share context across sessions without reloading a long journal or inventing answers to questions already settled. Stable architecture stays in
docs/; mutable working knowledge lives underkb-brain/.What
paad:kb-brainkb-brain/(no bulk ingest ofdocs/)minimal/standard/strict) with index-first retrievalSEAL.json(SHA-256); post-close corrections only viaamendments/scripts/kb_brain.py+ Make targets (kb-index,kb-new,kb-check, …)paad:brief-ruminatereview-neededapproved-specpushback,alignment, oragentic-reviewExplicit non-goals
1.24.1; changelog under[Unreleased](bump is a separate/release)Tests / validation
make export && make test(includesclaude plugin validate, digraph/help/readme/frontmatter/export checks, and newcheck-kb-brainunit tests covering init, workspace lifecycle, routing/IDs, seal/amend, and brief candidate-spec generation)