This hub gathers the process flows of the system — how knowledge gets in, how it gets out, and how the loop closes. On capture, the [JIRA integration concept] (Feature_JIRAIntegration) ties proactive knowledge capture to the existing dev process, with AI monitoring closed tickets and epics to prompt for undocumented rationale; the [knowledge-gap workflow] (Feature_KnowledgeGapWorkflow) closes the loop the other way — when a query fails, an expert is pinged to fill the hole and confirm it's filled. On the pipeline side, the [individuation and story-generation pipelines] (Narrative_SIC_Flows_1) lay out the stages: extract → diff → TX → user review → commit, with story generation as compile → template → model → output. The [end-to-end story flow] (Obs_StoryFlow) makes the runtime concrete — mcp/draft-story runs a .story file against a snapshot and returns Markdown synchronously. Sitting alongside (related, not core) is the [intentional-memory stance] (Observation_IntentionalMemory): memory chosen and committed deliberately rather than auto-summarized.
Witness phrases
verbatim AI monitors closed tickets/epics and prompts for undocumented knowledge: 'Hey, this epic—why did you implement it?'
verbatim When sales asks a question and collective memory doesn't know, ping the expert: 'Hey the customer's chief architect, you forgot to talk to AI about thing.' Then you do that and ping back: 'It's in there now, you're good.'
verbatim Individuation pipeline: extract → diff → TX → user review → commit. Evolved variant: snapshot + message → transaction. Story generation: compile → template → model → output.
verbatim One is to make memory intentional. So instead of this context squishing autosummarization, we have an agent or context that chooses to remember stuff, offers that to the user, and then saves that to the git repo intentionally
verbatim Intentional, user-offered memory vs automatic context auto-summarization — an explicit stance against the squishing default.
verbatim The individuation pipeline carries a superseded framing: original extract→diff→TX→user-review→commit vs an evolved 'snapshot + message → transaction' variant someone might still reach for.
verbatim Two opposite capture triggers — proactive monitoring of closed JIRA work vs reactive backfill on a failed query — same loop approached from opposite ends.
verbatim Story generation is synchronous (immediate Markdown return), which sits in tension with the multi-stage, review-gated individuation pipeline.
StyleObservationObservation_IntentionalMemory One is to make memory intentional. So instead of this context squishing autosummarization, we have an agent or context that chooses to remember stuff, offers that to the user, and then saves that to the git repo intentionally