| name | onboard |
| description | Use at the start of a session on an existing project, when joining a project, or the user says "catch me up", "what do I need to know", or "onboard me" |
Onboard to Project
Pull institutional knowledge from EC to get context on an existing project.
Announce: "I'm using the onboard skill to pull project context from EC."
The Flow
Verify EC → Load Config → Pull Memories → Summarize → Ready
Step 1: Verify EC Connection
ec_search: test connection
If EC unavailable: Stop. Cannot onboard without EC.
Step 2: Load Project Configuration
ec_search: project config
Extract:
- Test command
- Lint command
- Build command
- Branch convention
If no config found:
"This project hasn't been initialized for cogitation. Want me to set it up?"
If yes → Use @cog-init
Step 3: Pull All Memories
Search each memory type:
ec_search: type:decision
ec_search: type:learning
ec_search: type:pattern
Step 3.5: Graphify Structural Recon (STRICT when enabled)
Read .cogitation/config.json. If graphify.enabled is true, structural recon is REQUIRED — do not skip it.
- Ensure a graph exists (the AST build needs no model, no API key):
test -f graphify-out/graph.json || graphify update .
- Orient on structure:
graphify query "What are the main components and how do they connect?" --budget 800
- For specific symbols later, use
graphify query "..." / graphify explain "<symbol>".
The two sources are complementary: EC = decisions/gotchas/patterns; graphify = code structure/call-graph. Fold both into the briefing. If graphify.enabled is absent or false, skip this step entirely.
Step 4: Categorize and Summarize
Organize findings by area:
Architectural Decisions
List decisions with their rationale. Group by component/area.
Learnings (Gotchas & Workarounds)
List discovered issues and their solutions. Highlight anything that would be costly to rediscover.
Patterns (Conventions)
List established patterns and conventions specific to this project.
Step 5: Present Summary
Format as a briefing:
Project Context Loaded
Configuration:
- Test:
<command>
- Lint:
<command>
- Build:
<command>
Key Decisions (N):
- [Area]: [Decision summary]
- ...
Gotchas to Know (N):
- [Area]: [Learning summary]
- ...
Conventions (N):
Ready to work. What would you like to do?
Step 6: Offer Next Steps
{
"questions": [{
"question": "What would you like to do?",
"header": "Next",
"options": [
{ "label": "Start a feature", "description": "Use @brainstorming" },
{ "label": "Fix a bug", "description": "Use @debugging" },
{ "label": "Review memories", "description": "Use @audit" },
{ "label": "Just exploring", "description": "I'll ask when ready" }
],
"multiSelect": false
}]
}
When to Use
- New session - At start of conversation to load context
- After break - Returning to project after time away
- New team member - Onboarding someone unfamiliar with decisions
- Context refresh - When you need to remember what was decided
What Makes Good Onboarding
| Do | Don't |
|---|
| Summarize, don't dump | List every memory verbatim |
| Highlight gotchas prominently | Bury important warnings |
| Group by area/component | Random ordering |
| Note what's missing | Assume completeness |
