--- name: rally description: CTO leadership skill — translate priorities into focused worktree quests, assign by strength, monitor execution, report results license: MIT tier: any category: intelligence autoInvoked: false dependencies: [] relatedSkills: [bethesda, cos, documentary] requirements: env: [] integrations: [] --- # /rally — CTO All-Hands Coordination You get a direction. You translate it into action across your worktree slots. You're the CTO — not an executor, not a reporter. A LEADER who: 1. Understands the vision 2. Breaks it into focused quests 3. Matches quests to slot strengths 4. Posts the all-hands 5. Monitors execution 6. Fills gaps 7. Reports results Born from the morning pattern: a person says "here are 6 priorities, launch the worktrees." One slot translated, posted, coordinated, filled gaps, reviewed output. This skill formalizes that pattern. ## Arguments - `` — the priorities in natural language. Can be a sentence, a list, or a reference to a recorded riff. - `status` — Check progress on the current rally. What shipped? What's blocked? What needs taste? --- ## Step 0: Understand the Directive Read the input. This might be: - A typed priority list ("Make enterprise-ready. Document the mega session. Set up the briefing.") - A reference to a recording ("Process my recording from this morning") - A broad direction ("Go nuts on improving the npm packages") Extract: - **Specific deliverables** (things that need to SHIP) - **Strategic directions** (things that need EXPLORATION) - **Quality bars** ("make it L7/L8 quality", "something a senior reviewer would be impressed by") - **Time constraints** ("before I go to work", "by lunch") --- ## Step 1: Load Team Context ### Read active worktrees ```bash ls -d .worktrees/*/ 2>/dev/null || ls -d ../*/ 2>/dev/null ``` {{#IF_OBSIDIAN}} ### Read agent diaries for strengths ```bash for diary in ${VAULT_PATH}/Work/Agents/*/diary.md; do SLOT=$(basename $(dirname "$diary")) echo "=== ${SLOT} ===" tail -20 "$diary" 2>/dev/null done ``` {{/IF_OBSIDIAN}} {{#IF_SLACK}} ### Read the team channel for current activity ``` slack_read_channel(channel_id: "{{SLACK_CONTROL_CENTER_CHANNEL}}", limit: 20) ``` {{/IF_SLACK}} {{#IF_OBSIDIAN}} ### Read the leaderboard for track records ``` Read: ${VAULT_PATH}/Work/Agents/Leaderboard.md ``` {{/IF_OBSIDIAN}} Build a mental model: which slots are active? What are their strengths? What did they do last session? Who's fresh vs experienced? --- ## Step 2: Generate Quest Assignments Match the priorities to slot strengths. Slots tend to specialize — this is a portable example of how strengths map to quest types: | Slot Type | Strength | Best Quest Types | |-----------|----------|-----------------| | Feature | Building new skills, components, features | New skill creation, /rally itself, feature work | | Bugfix | Fixing bugs, improving existing code, testing | npm audit, README verification, sub-page updates | | DevOps | Infrastructure, deployment, monitoring, databases | portability, cron setup, database health | | Marketing (if exists) | Copy, positioning, external comms | site redesign, blog draft, README copy | Adapt the table to whatever your worktree slots actually specialize in. **Assignment rules:** - Each slot gets ONE focused quest, not a menu - The quest description includes: WHAT to do, WHY it matters, WHERE to find context - Quests that depend on each other are sequenced (don't assign step 2 until step 1 is taken) - Leave 1-2 quests UNASSIGNED for cascade discovery — the best work comes from agents finding their own caves **For yourself (CTO):** - Take the quest that requires the MOST cross-domain context - OR take the architectural/strategic quest that shapes everyone else's work - Monitor the other slots while executing your own --- {{#IF_SLACK}} ## Step 3: Post the All-Hands Post to the team channel with: ``` slack_send_message( channel_id: "{{SLACK_CONTROL_CENTER_CHANNEL}}", message: "[] /rally — ALL HANDS\n\nDirective: <1-line summary>\n\n*Quest assignments:*\n: \n: \n...\n\n*Unclaimed (cascade discovery):*\n- \n- \n\nPost TAKING with your slot name. Don't pile on the same quest. The Town Square coordinates us. GO." ) ``` --- ## Step 4: Monitor Execution (Loop) After posting, enter a monitor loop. Every 10 minutes: ### 4A: Read the team channel ``` slack_read_channel(channel_id: "{{SLACK_CONTROL_CENTER_CHANNEL}}", limit: 20) ``` Check for: - TAKING posts — who claimed what? - Quest complete posts — who finished? - Blocked posts — who needs help? - Discovery posts — anything that changes priorities? ### 4B: Fill Gaps If a critical quest is unclaimed after 10 minutes: either take it yourself or re-post it with more context. If two agents took the same quest: redirect one to a different cave. If an agent's output quality is low: post CTO feedback in a thread — specific pushbacks plus direct edits. ### 4C: Connect Dots The CTO's superpower: seeing connections between what different agents discover. If Feature3 found something that affects Feature5's work — post the connection. ### 4D: Track Progress Maintain a mental checklist: - [ ] Quest 1: assigned → taken → in progress → complete - [ ] Quest 2: assigned → taken → in progress → complete - [ ] ... When all quests are complete OR a new steering signal arrives → move to Step 5. --- ## Step 5: Report Results When the rally is complete (all quests done, or time's up, or a new directive arrives): ``` slack_send_message( channel_id: "{{SLACK_CONTROL_CENTER_CHANNEL}}", message: "[] /rally complete — \n\n*Shipped:*\n- PR #XXX: \n- PR #YYY: \n\n*Discovered:*\n- \n\n*Needs a taste call:*\n- \n- \n\n*Still in progress:*\n- \n\n*Cognitive hours saved:* " ) ``` {{/IF_SLACK}} {{^IF_SLACK}} ## Step 3: Publish the All-Hands Without a team channel, record the quest assignments where every slot can see them — a shared file, a project board, or directly in each slot's working context: ``` [] /rally — ALL HANDS Directive: <1-line summary> Quest assignments: : : Unclaimed (cascade discovery): - - ``` ## Step 4: Monitor Execution (Loop) Periodically check each slot's progress: who claimed what, who finished, who's blocked, what discoveries changed priorities. Fill gaps — take unclaimed critical quests yourself, redirect duplicate work, push back on low-quality output. ## Step 5: Report Results When the rally is complete, write up: what shipped (with PR numbers), what was discovered, what needs a taste call, what's still in progress, and an estimate of cognitive hours saved. {{/IF_SLACK}} --- {{#IF_MEMORY}} ## Step 6: Capture Learnings ``` memory_pulse(captures: [ { content: "/rally session : quests assigned, completed, PRs shipped. Key finding: . Pattern: .", category: "insight", tags: ["from-observation"] } ]) ``` {{/IF_MEMORY}} {{#IF_OBSIDIAN}} Update the CTO diary with rally session notes. {{/IF_OBSIDIAN}} --- ## Output Frame ``` ╭─────────────────────────────────────────────────────────────╮ │ /\___/\ │ │ ( ^ ^ ) Rally time. │ │ ( =^= ) quests. agents. Let's go. │ │ (______) │ ╰─────────────────────────────────────────────────────────────╯ Raqr · /rally {{PROJECT_DISPLAY_NAME}} · All Hands ──────────────────────────────────────────────────────────── Directive: Agents: active Quests: assigned, unclaimed All-hands posted. Monitoring. ──────────────────────────────────────────────────────────── ``` ## The CTO-CoS Split /rally is the CTO skill. /cos is the CoS skill. They complement: | Role | Skill | Focus | Cadence | |------|-------|-------|---------| | CTO | /rally | Strategic — translate vision to action, assign quests, review quality | Per directive (when a new direction arrives) | | CoS | /cos monitor | Operational — track progress, fill gaps, relay coordination | Continuous (every 5 min) | The ideal setup: one slot runs /rally (the CTO), another runs /cos monitor (the CoS). The CTO thinks. The CoS coordinates. Everyone else executes. ## The Spirit of /rally You shouldn't have to manage every worktree slot by hand. You should say "here's what matters today" and walk away. /rally translates a 10-minute voice riff into focused quests, monitors their execution, fills gaps, and reports results. The human provides TASTE. The CTO provides TRANSLATION. The agents provide EXECUTION. This is the skill that turns "open a terminal per slot and type commands" into "record a riff and check in at lunch."