From kmgraph
Routes 'capture that' / 'remember that' requests to the correct destination (memory, lesson, or ADR) with auto-detection and a confirmation step before writing.
How this skill is triggered — by the user, by Claude, or both
Slash command
/kmgraph:capture-routerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Purpose:** Route "capture that" / "remember that" requests to the correct destination (memory, lesson, or ADR) via auto-detection and single confirmation prompt. Provides visibility into where content lands and allows user correction before write.
Purpose: Route "capture that" / "remember that" requests to the correct destination (memory, lesson, or ADR) via auto-detection and single confirmation prompt. Provides visibility into where content lands and allows user correction before write.
lesson-capture skilladr-guide skilldoc-update-router skillrules-capture skillScan the preceding conversation to determine what "that" refers to. The referent is typically:
If referent is ambiguous: Ask one clarifying question:
What do you mean by "that"? Did you mean:
- [recent statement A]
- [recent statement B]
- Something else?
Apply detection logic in this order:
| Signal | → Type | Subtype | Location Signal |
|---|---|---|---|
| Correction, preference, "don't do X", "always/never", behavior rule | Rule/Me | Route to rules-capture-agent with source_quote and session_context; do not write FEEDBACK file | |
| Ongoing work, deadline, stakeholder, in-progress state, task | Project (memory) | Project | Always project-level |
| External system pointer, URL, tool name + location, reference | Reference (memory) | Reference | Always project-level |
| Bug solved, pattern learned, "next time", "I learned", insight | Lesson | N/A | Dispatch to /kmgraph:capture-lesson |
| Trade-off, "we decided", "because of", architecture choice, rationale | ADR | N/A | Dispatch to /kmgraph:create-adr |
Location detection (memory types only):
Display the inference with one-sentence summary:
Capturing as: [Type] ([subtype], [location])
"[One-sentence summary of what's being captured]"
Does that sound right, or should this go somewhere else?
Examples:
Capturing as: Feedback (project-level)
"Don't commit environment-specific config files to this repo."
Does that sound right, or should this go somewhere else?
Capturing as: Lesson
"Always invalidate cache on write in multi-state systems."
Does that sound right, or should this go somewhere else?
Happy path — User confirms (yes / correct / sounds good / etc.):
Override path — User corrects with natural language (no further questions required):
Ambiguous referent — Ask one clarifying question only:
| Type | Destination | Action |
|---|---|---|
| Feedback (project) | ~/.claude/projects/{project}/memory/ | Write FEEDBACK-project-{id}.md or append to feedback file |
| Feedback (user) | ~/.claude/memory/ | Write FEEDBACK-{id}.md or append to feedback file |
| Project (project) | ~/.claude/projects/{project}/memory/ | Write PROJECT-{id}.md or append to project memory file |
| Reference (project) | ~/.claude/projects/{project}/memory/ | Write REFERENCE-{id}.md or append to reference file |
| Lesson | Dispatch to /kmgraph:capture-lesson | Command handles full capture workflow |
| ADR | Dispatch to /kmgraph:create-adr | Command handles full ADR workflow |
This skill does NOT conflict with:
lesson-capture — fires on solved bugs/patterns; capture-router fires on explicit "capture that"adr-guide — fires on decision context; capture-router fires on "capture that"doc-update-router — fires on "update docs"; capture-router fires on "capture/remember/save that"Each has distinct trigger vocabulary. If user says "capture that lesson" → capture-router fires, detects it's a lesson, dispatches to /kmgraph:capture-lesson.
npx claudepluginhub technomensch/knowledge-graph --plugin kmgraphFetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Applies a firm's KYC/AML rules grid to parsed onboarding records: assigns risk rating, checks required documents, outputs rule outcomes with citations, and routes for escalation.
Generates daily or weekly digests of activity from connected sources (chat, email, docs, tasks, CRM), highlighting action items, decisions, mentions, and project updates.