From kmgraph
Prompts for session summary when stopping work, reaching context limits, or completing milestones. Flags mid-plan items, open decisions, and uncaptured lessons.
How this skill is triggered — by the user, by Claude, or both
Slash command
/kmgraph:session-wrapThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Purpose:** Prompt for session summary when user indicates they're stopping work, approaching context limits, reaching milestones, or when open plan items, draft decisions, or uncaptured lessons are detected.
Purpose: Prompt for session summary when user indicates they're stopping work, approaching context limits, reaching milestones, or when open plan items, draft decisions, or uncaptured lessons are detected.
Trigger Indicators:
User Intent Signals:
Active Work Signals:
docs/plans/*.md contains unchecked - [ ] items (mid-plan indicator)knowledge/decisions/*.md has ADRs with Status: Proposed or Status: Draft (open decisions)fix, solved, implement, pattern, debug, refactor) but no corresponding lesson file in knowledge/lessons-learned/knowledge/rules.md or knowledge/me.md this session: surface "You established N guideline(s) this session. Worth checking whether any of them only apply in specific situations."Block Conditions:
/tmp/.kg-session-summarized-{kg-name}-{date} — if present, do NOT prompt (Stop hook already fired; avoid double-prompting)ECC Compatibility Note: This skill is coordinated with the Stop hook to prevent duplicate prompting. On Claude Code, the /tmp flag mechanism is used for inter-process coordination. On other ECC platforms, equivalent synchronization mechanisms (environment variables, shared state, or hook phase detection) should be used to prevent re-prompting if the Stop hook has already triggered this workflow.
Snapshot Awareness:
/tmp/.kg-snapshot-{YYYY-MM-DD} (today's date)"You took a session snapshot earlier — want to complete the wrap-up and save the final summary?"
Behavior:
When triggered, directly dispatch to session-summary-agent with conversational language that addresses the user, never exposing internal mechanics.
Example Triggers:
User: "Alright, I've pushed this to the branch. Wrapping up for today."
→ "Before you go — want a quick note on what we worked on today?"
[Skill detects unchecked plan steps]
→ "You're mid-plan — want to mark off what we finished before wrapping up?"
[Skill detects lesson-worthy commits without corresponding lesson]
→ "Looks like some meaningful work wasn't captured — want to save a note before finishing?"
User-Facing Language Rules:
After writing the session summary, scan all ADRs and ENHs created or modified this session:
git diff --name-only HEAD~20 -- knowledge/decisions/ knowledge/enhancements/ 2>/dev/null \
| head -20
For each file found, extract the ## Open Questions section (if present).
Deduplicate across all files (same question text = one entry).
Emit a final ## Open Items section in the session summary with all extracted questions.
Single write path enforced: Do NOT write open items directly to the session summary during the session. Only the session-wrap skill reads and aggregates them at wrap time.
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.