| name | nexus |
| description | Use for persistent TODO capture, daily brief retrieval, and shared Codex/Claude operating rules. Trigger on remember/track/follow-up requests, "nexus add/show/brief", or policy/workflow questions across Slack, Outlook, and Jira contexts. When in doubt, use this skill.
|
Nexus Agent Kit
Shared operating rules for Codex and Claude Code sessions.
Compatibility
- The plugin ships its own dependency-free scripts: run them with
node "${CLAUDE_PLUGIN_ROOT}/bin/nexus.js" (requires Node >= 18; no npm install)
- Required files:
~/.nexus/TODOS.md, ~/.nexus/state.json
Workflow
Step 1 — TODO Classification & Addition
When the user asks to remember, track, add, or follow up on something:
- Classify into exactly one of:
Work, Python, Claude, Codex, AWS, Jira, Slack, Outlook, Personal, General.
- Run
node "${CLAUDE_PLUGIN_ROOT}/bin/nexus.js" add "<task>".
- Confirm with the output format below.
If the script fails: check that Node >= 18 is installed (node --version).
Step 2 — TODO Listing
Run node "${CLAUDE_PLUGIN_ROOT}/bin/nexus.js" todos --limit 8. Output grouped by label, sorted by recency.
Step 3 — Daily Brief
When the user asks for a daily brief or "what's on my plate":
- Read
~/.nexus/TODOS.md.
- Read
~/.nexus/state.json.
- Use Slack, Outlook, and Jira MCPs only if configured in state.json.
- Extract action items and add each with
node "${CLAUDE_PLUGIN_ROOT}/bin/nexus.js" add "<task>".
- Ask before sending Slack messages, email, updating Jira, or changing AWS.
Output Format
For TODO operations:
Outcome: [what was done]
Actions: [commands run]
Evidence: [nexus script output or confirmation]
Next Step: [what the user should do next]
For daily briefs:
Outcome:
Urgent:
Today:
Waiting:
Added TODOs:
Risks:
Next Step:
Anti-Patterns
- Never perform external writes (Slack, email, Jira, AWS changes) without explicit user confirmation first.
- Never expose API keys, tokens, or secrets in any output.
- Never skip classification — always assign a label before adding a TODO.
- Never use a large model for TODO classification or text summarization; use the smallest capable model.
- Never claim nexus script output without actually running the command.
Examples
Input: "Remember to follow up with finance about the Outlook invoice."
Output:
Outcome: TODO added under Work.
Actions: node "${CLAUDE_PLUGIN_ROOT}/bin/nexus.js" add "Follow up with finance about Outlook invoice"
Evidence: ✓ Added to ~/.nexus/TODOS.md
Next Step: Review during tomorrow's daily brief.
Input: "What's on my plate today?"
Output:
Outcome: Daily brief generated from TODOS.md and state.json.
Urgent: Deploy hotfix for auth service (due today)
Today: Review PR #42, sync with design team
Waiting: Finance invoice reply
Added TODOs: none new
Risks: Auth deploy blocks two downstream services
Next Step: Start with the auth hotfix deploy.
Safety Rules
- Prefer read-only MCP operations in all cases.
- Ask before any external write — once per action, not once per session.
- Verify command output before claiming a task is complete.