davepoon/ops-yolo

Runtime Context

Before YOLO analysis, load:

1. Preferences: cat ${CLAUDE_PLUGIN_DATA_DIR:-$HOME/.claude/plugins/data/ops-ops-marketplace}/preferences.json — read owner, timezone, yolo_enabled, all channel configs
2. Daemon health: cat ${CLAUDE_PLUGIN_DATA_DIR}/daemon-health.json — all services must be healthy for comprehensive analysis
3. Secrets: Resolve ALL keys via env → Doppler → password manager: GITHUB_TOKEN, SENTRY_AUTH_TOKEN, LINEAR_API_KEY, AWS_ACCESS_KEY_ID
4. Ops memories: Load ALL files from ${CLAUDE_PLUGIN_DATA_DIR}/memories/ — contact profiles, preferences, topics, donts. YOLO agents need maximum context.

OPS ► YOLO MODE

CLI/API Reference

aws CLI (Cost Explorer)

| Command | Usage | Output | |---------|-------|--------| | aws ce get-cost-and-usage --time-period Start=<YYYY-MM-DD>,End=<YYYY-MM-DD> --granularity MONTHLY --metrics "UnblendedCost" --output json | Current month spend | Cost JSON |

gh CLI (GitHub)

| Command | Usage | Output | |---------|-------|--------| | gh pr list --repo <owner/repo> --json number,title,statusCheckRollup,reviewDecision,mergeable,isDraft | Open PRs with status | JSON array | | gh pr merge <n> --repo <repo> --squash --admin | Squash merge PR | Merge result | | gh run list --limit 20 --json status,conclusion,name,headBranch,createdAt | Recent CI runs | JSON array |

---

Agent Teams support

If CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set, use Agent Teams instead of fire-and-forget subagents for the C-suite analysis (Phase 2). This enables:

• Agents can share findings mid-analysis (CEO discovers a revenue blocker → CFO factors it into ROI)
• You can steer agents if early findings change priorities
• Agents coordinate on the consensus recommendation

Team setup (only when flag is enabled):

TeamCreate("yolo-csuite")
Agent(team_name="yolo-csuite", name="ceo", subagent_type="ops:yolo-ceo", ...)
Agent(team_name="yolo-csuite", name="cto", subagent_type="ops:yolo-cto", ...)
Agent(team_name="yolo-csuite", name="cfo", subagent_type="ops:yolo-cfo", ...)
Agent(team_name="yolo-csuite", name="coo", subagent_type="ops:yolo-coo", ...)

After initial analysis, use SendMessage(to="cto", content="CFO flagged $400/mo in waste — does this change your tech-debt ranking?") or similar to cross-pollinate findings between peer agents. The main /ops:yolo orchestrator (this skill) then reads all four analysis files (ceo-analysis.md, cto-analysis.md, cfo-analysis.md, coo-analysis.md) and synthesizes them into the Hard Truths report. yolo-ceo is a parallel peer, not the synthesizer.

If the flag is NOT set, fall back to standard parallel subagents (fire-and-forget, no mid-task steering).

Phase 1 — Pre-gather ALL data

Run all of these simultaneously:

${CLAUDE_PLUGIN_ROOT}/bin/ops-infra 2>/dev/null || echo '{}'

${CLAUDE_PLUGIN_ROOT}/bin/ops-git 2>/dev/null || echo '[]'

${CLAUDE_PLUGIN_ROOT}/bin/ops-prs 2>/dev/null || echo '[]'

${CLAUDE_PLUGIN_ROOT}/bin/ops-ci 2>/dev/null || echo '[]'

${CLAUDE_PLUGIN_ROOT}/bin/ops-unread 2>/dev/null || echo '{}'

aws ce get-cost-and-usage --time-period "Start=$(date +%Y-%m-01),End=$(date +%Y-%m-%d)" --granularity MONTHLY --metrics "UnblendedCost" --output json 2>/dev/null || echo '{}'

cat "${CLAUDE_PLUGIN_ROOT}/scripts/registry.json" 2>/dev/null || echo '{}'

${CLAUDE_PLUGIN_ROOT}/bin/ops-external 2>/dev/null || echo '[]'

for d in $(jq -r '.projects[] | select(.gsd == true) | .paths[]' "${CLAUDE_PLUGIN_ROOT}/scripts/registry.json" 2>/dev/null); do
  expanded="${d/#\~/$HOME}"
  [ -f "$expanded/.planning/STATE.md" ] && echo "=== $(basename $expanded) ===" && cat "$expanded/.planning/STATE.md" && echo "---"
done

---

Phase 2 — Spawn 4 C-suite agents in parallel

Spawn these 4 agents simultaneously using all pre-gathered data as context. Each writes their analysis to a file in /tmp/yolo-[session]/:

Agent 1 — CEO (Strategic)

Uses agents/yolo-ceo.md. Writes /tmp/yolo-[session]/ceo-analysis.md.

• What's the #1 thing blocking growth right now?
• Are we building the right things?
• Where are we wasting time vs. creating value?
• What would you tell an investor today, unfiltered?

Agent 2 — CTO (Technical)

Uses agents/yolo-cto.md. Writes /tmp/yolo-[session]/cto-analysis.md.

• What's the worst technical debt that will bite us?
• Which services are time-bombs?
• Is the team/architecture set up to scale?
• What corners were cut that need fixing now?

Agent 3 — CFO (Financial)

Uses agents/yolo-cfo.md. Writes /tmp/yolo-[session]/cfo-analysis.md.

• Actual burn rate vs. runway
• Which AWS services are waste?
• When do we hit zero if nothing changes?
• What's the ROI on current work?

Agent 4 — COO (Operations)

Uses agents/yolo-coo.md. Writes /tmp/yolo-[session]/coo-analysis.md.

• What's falling through the cracks right now?
• Which processes are broken?
• What's the top execution risk this week?
• What should be automated that isn't?

---

Phase 3 — Hard Truths Report (orchestrator synthesis)

This skill (the main orchestrator) is the synthesizer — NOT yolo-ceo. After all 4 parallel agents complete and have written their analysis files to /tmp/yolo-[session]/{ceo,cto,cfo,coo}-analysis.md, read all four files here in the main context and synthesize them into a unified report:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 YOLO ► HARD TRUTHS REPORT — [date]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

 CEO: [1-2 brutal strategic truths]

 CTO: [1-2 brutal technical truths]

 CFO: [1-2 brutal financial truths]

 COO: [1-2 brutal operational truths]

──────────────────────────────────────────────────────
 CONSENSUS: The #1 thing that matters today is:
 [single most important action, no sugar-coating]
──────────────────────────────────────────────────────

 Full analysis files saved to:
 /tmp/yolo-[session]/ceo-analysis.md
 /tmp/yolo-[session]/cto-analysis.md
 /tmp/yolo-[session]/cfo-analysis.md
 /tmp/yolo-[session]/coo-analysis.md

──────────────────────────────────────────────────────
 Type YOLO to hand over the controls.
 I'll run your business autonomously for the next day.
 This means: closing inbox, merging ready PRs,
 fixing fires, advancing GSD phases, triaging issues.

 Or pick an analysis to read:
──────────────────────────────────────────────────────

Use batched AskUserQuestion calls (max 4 options each):

AskUserQuestion call 1:

  [Read CEO analysis]
  [Read CTO analysis]
  [Read CFO analysis]
  [More...]

AskUserQuestion call 2 (only if "More..."):

  [Read COO analysis]
  [Execute top recommendation now]
  [Type YOLO to go autonomous]

---

Phase 4 — YOLO Autonomous Mode

If user types YOLO (all caps), enter autonomous mode via /loop.

Before starting, use AskUserQuestion to confirm scope:

YOLO mode will autonomously execute these steps:
  1. Inbox — reply to humans, archive automated
  2. Fires — fix CRITICAL/HIGH production issues
  3. PRs — merge ready PRs (CI green, approved)
  4. Triage — auto-resolve confirmed-fixed issues
  5. GSD — advance highest-priority phase
  6. Linear — sync sprint board
  7. Deploy — trigger pending deploys
  8. Report — summary

  [Run all 8 steps]  [Pick which steps to run]  [Cancel]

If user picks "Pick which steps", show steps as multiSelect via batched AskUserQuestion calls (max 4 options each):

Call 1: [Inbox], [Fires], [PRs], [More steps...] Call 2 (if "More steps..."): [Triage], [GSD], [Linear], [More steps...] Call 3 (if "More steps..."): [Deploy], [Report], [Done selecting]

Run the selected steps in sequence, reporting after each step.

Per-step confirmations (use AskUserQuestion before EACH destructive action):

• Inbox: Show drafted replies and ask [Send all N replies] / [Review each one] / [Skip inbox] before sending any messages
• Fires: Show proposed fix and ask [Dispatch fix agent] / [Skip] before each agent dispatch
• PRs: Show PR list and ask [Merge all N ready PRs] / [Pick which ones] / [Skip] before merging
• Triage: Show issues to close and ask [Auto-resolve all N confirmed-fixed] / [Review each] / [Skip] before closing
• Deploy: Show pending deploys and ask [Deploy all] / [Pick which] / [Skip] before triggering
• Infrastructure changes: For EVERY destructive infra action (delete ALB, stop RDS, disable Multi-AZ, purge images, etc.), present the specific action with context from the C-suite reports and ask [Execute] / [Skip] individually. NEVER batch destructive infra actions.

Report-driven execution: When the user approves executing recommendations from the Hard Truths report:

1. Read ALL C-suite analysis files (/tmp/yolo-[session]/*.md)
2. Extract specific actionable items marked with ⚠️ REQUIRES CONFIRMATION
3. Present each action individually via AskUserQuestion with the exact command that will run, the expected outcome, and the source report (CTO/CFO/COO)
4. Only execute after explicit per-action approval
5. After each action, verify the result and report back before proceeding to the next

After each step, check if new fires have appeared before proceeding. Report final summary when done.

If $ARGUMENTS is analyze or empty, go straight to Phase 1. If $ARGUMENTS is YOLO, skip to Phase 4. If $ARGUMENTS is report, skip to Phase 3 (reads existing analysis files if present).

---

Native tool usage

Tasks — progress tracking

Use TaskCreate at the start of Phase 4 to create a task for each YOLO step. Update with TaskUpdate as each completes. This gives the user a live progress view across the autonomous run.

PlanMode — review before autonomous

Before Phase 4 execution, use EnterPlanMode to present the full execution plan. The user reviews what YOLO will do, approves or modifies, then ExitPlanMode to begin execution.

Cron — schedule daily YOLO

After Phase 4 completes, offer to schedule recurring YOLO via AskUserQuestion:

  [Schedule daily YOLO at 9am]  [Schedule weekly Monday briefing]  [No schedule]

Use CronCreate if selected. Use CronList/CronDelete to manage existing schedules.

Monitor — live CI/deploy watching

When YOLO dispatches fix agents or triggers deploys, use Monitor to stream CI output in real-time instead of polling with sleep loops.

WebFetch/WebSearch — enrichment

Use WebFetch to pull Grafana dashboards, Sentry event details, or AWS status pages when MCPs are unavailable. Use WebSearch to find context on production errors (e.g., known AWS outages).