Agents Reference

Agents Reference

claude-ops ships 9 autonomous agents. They are spawned by skills and run as background subagents with their own tool access and turn budgets. Most are read-only; the exceptions are documented below.

---

comms-scanner

File: agents/comms-scanner.md
Model: claude-sonnet-4-5
Effort: low
Max turns: 10
Read-only: Yes (Write and Edit are disallowed)

Purpose: Scans all communication channels for full inbox state. Classifies each conversation as NEEDS_REPLY, WAITING (on other party), or HANDLED. Returns structured JSON.

Tools available: Bash, Read

What it does:

1. Runs WhatsApp scan via wacli chats list --json, checks last message sender on each chat to determine reply needed
2. Scans email inbox via gog gmail search "in:inbox" (up to 30 threads)
3. Reads Slack channel data if MCP is configured
4. Reads Telegram dialogs if Telegram MCP is configured
5. Classifies each conversation and returns structured output

Spawned by: /ops:inbox, /ops:go (for unread count summary)

---

infra-monitor

File: agents/infra-monitor.md
Model: claude-sonnet-4-5
Effort: low
Max turns: 15
Read-only: Yes (Write and Edit are disallowed)

Purpose: ECS, Vercel, and AWS health checker. Returns structured JSON with service health, recent deployments, and anomaly flags.

Tools available: Bash, Read, Vercel MCP (list_projects, list_deployments, get_deployment, get_runtime_logs)

What it does:

1. Lists all ECS clusters and checks running vs. desired task counts for each service
2. Checks stopped task reasons to identify crash-loops
3. Queries Vercel for recent deployment status per project
4. Flags services with health anomalies (mismatched counts, recent failures)

Spawned by: /ops:fires, /ops:deploy

---

project-scanner

File: agents/project-scanner.md
Model: claude-sonnet-4-5
Effort: low
Max turns: 15
Read-only: Yes (Write and Edit are disallowed)

Purpose: Git, PR, and CI status scanner across all registered repos. Returns structured JSON with branch state, uncommitted files, open PRs, and CI status.

Tools available: Bash, Read, Glob

What it does:

1. Loads scripts/registry.json to get all project paths and repos
2. Per project: checks current branch, dirty file count, ahead/behind counts
3. Queries GitHub via gh for open PR count per repo
4. Checks last GitHub Actions run status per repo

Spawned by: /ops:projects, /ops:go

---

revenue-tracker

File: agents/revenue-tracker.md
Model: claude-sonnet-4-5
Effort: medium
Max turns: 20
Read-only: Yes (Write and Edit are disallowed)

Purpose: Revenue, billing, and credits analysis. Queries AWS Cost Explorer, checks credit balances, cross-references project revenue stages.

Tools available: Bash, Read

What it does:

1. Queries AWS Cost Explorer for current month costs by service
2. Pulls last 3 months trend data
3. Checks AWS credits balance and expiry dates
4. Cross-references project revenue.stage and revenue.mrr from registry
5. Calculates burn rate and estimated runway

Spawned by: /ops:revenue

---

triage-agent

File: agents/triage-agent.md
Model: claude-sonnet-4-5
Effort: high
Max turns: 40
Read-only: No — can create fix branches and PRs

Purpose: Investigates a specific issue from Sentry, Linear, or GitHub. Finds the root cause, checks if already fixed, and either confirms resolution or creates a fix branch with a PR.

Tools available: Bash, Read, Grep, Glob, Edit, Write, Linear MCP (get_issue, save_issue, save_comment)

Input provided by calling skill:

• ISSUE_SOURCE: sentry | linear | github
• ISSUE_ID: issue identifier
• ISSUE_TITLE: human-readable title
• ISSUE_BODY: full description and stack trace
• AFFECTED_REPO: which repo to investigate
• AFFECTED_FILE: optional file from stack trace

What it does:

1. Searches codebase for the error pattern (grep by error message, class name, file)
2. Checks if error is already fixed in recent commits
3. If fixed: marks issue as resolved in Linear/Sentry and adds comment
4. If not fixed: investigates root cause, writes fix, creates branch and PR
5. Updates Linear issue with findings and PR link

Spawned by: /ops:triage, /ops:fires (when dispatching fix agent)

---

yolo-ceo

File: agents/yolo-ceo.md
Model: claude-opus-4-6
Effort: high
Max turns: 20
Read-only: No (can write analysis files to /tmp/yolo-[session]/)

Purpose: Strategic priority synthesis. The final synthesizer in the YOLO C-suite pipeline. Reads CTO, CFO, and COO reports, resolves conflicts between their recommendations, and produces the executive summary.

Tools available: Bash, Read, Write, Grep, Glob, Linear MCP (list_issues, list_cycles, list_projects)

Reporting chain: Runs after CTO, CFO, and COO complete. Reads their analysis files from /tmp/yolo-[session]/. Is the only agent that presents output to the user.

Analysis covers:

• Growth blockers (#1 thing blocking growth)
• Resource allocation (build vs. buy decisions)
• What to tell an investor, unfiltered
• Resolution of CFO/CTO conflicts (cut cost vs. invest in tech)

Spawned by: /ops:yolo

---

yolo-cto

File: agents/yolo-cto.md
Model: claude-opus-4-6
Effort: high
Max turns: 25
Read-only: No (writes to /tmp/yolo-[session]/cto-analysis.md)

Purpose: Technical health analysis. Identifies the specific time-bombs in the codebase, services at risk, scalability limits, and cut corners.

Tools available: Bash, Read, Write, Grep, Glob

Analysis covers:

• Worst tech debt that will cause a production incident or 2-week rewrite
• Which service is one bad deploy away from a P0
• What breaks first at 10x current load (database, API, auth, queue)
• What corners were cut that need fixing now

Spawned by: /ops:yolo (in parallel with CFO and COO)

---

yolo-cfo

File: agents/yolo-cfo.md
Model: claude-opus-4-6
Effort: high
Max turns: 20
Read-only: No (writes to /tmp/yolo-[session]/cfo-analysis.md)

Purpose: Financial analysis. Follows the money without optimism bias.

Tools available: Bash, Read, Write

Analysis covers:

• Actual burn rate vs. perceived (real AWS numbers)
• Which services are waste (orphaned resources, unused capacity)
• Runway calculation based on current trajectory
• ROI on current engineering work
• Credits expiry and impact on runway

Spawned by: /ops:yolo (in parallel with CTO and COO)

---

yolo-coo

File: agents/yolo-coo.md
Model: claude-opus-4-6
Effort: high
Max turns: 25
Read-only: No (writes to /tmp/yolo-[session]/coo-analysis.md)

Purpose: Operations execution analysis. Finds what the CEO doesn't see — work falling through the cracks, broken processes, communication failures.

Tools available: Bash, Read, Write, Grep, Glob, Linear MCP (list_issues, list_cycles)

Analysis covers:

• PRs open for more than 7 days (stale review pipeline)
• Linear issues with no activity in 5+ days
• GSD phases stuck at the same progress for over a week
• Repos with no commits in 2+ weeks (abandoned work)
• Messages in channels with no response (communication failures)

Spawned by: /ops:yolo (in parallel with CTO and CFO)