PM Agent Manual

1. What the PM Agent does

The PM Agent is your always-on project manager. It connects to your ticketing system (Linear, GitHub Issues, Jira) and keeps track of every open item — assigning priorities, flagging stale tickets, sending SLA alerts before breaches happen, and generating weekly health digests for your team.

It also owns your incident workflow end-to-end: detecting severity, paging the right people, tracking resolution progress, and writing the postmortem when the dust settles. You never have to remind it of a follow-up — it tracks all open threads and follows up automatically.

What it connects to

2. Tracking tickets and work items

Give the agent access to your project board by sharing the connection details in your first message. It then monitors every open ticket and maintains a live health picture of your backlog.

How to connect your ticketing system

What the agent tracks per ticket

Staleness detection

Staleness means a ticket has not been updated in longer than your configured threshold. The default is 5 days for P2 and below, 24 hours for P1, and 4 hours for P0. The agent sends a Slack/email alert for each stale ticket and pings the assignee if one is set.

3. Incident management

The PM Agent manages the entire lifecycle of an incident — from declaration through to postmortem. When an incident is declared (by you, by a monitoring alert, or by the DevOps Agent), the PM Agent takes over coordination.

Incident lifecycle

Declaring an incident

Severity definitions

4. SLA monitoring

The PM Agent enforces your service-level agreements automatically. You define the SLA rules once; the agent monitors every ticket against them and alerts you before a breach — not after.

How to define SLA rules

SLA alert schedule

Reading the SLA dashboard

Ask the agent for a live SLA status at any time: "What is our current SLA compliance rate this month?" The agent responds with: tickets tracked, tickets breached, breach rate by priority, average resolution time, and the tickets currently at risk.

5. Sprint health and digests

The PM Agent sends a weekly digest every Monday morning (configurable) covering the health of your sprint, backlog, and team. It also sends a briefer daily standup summary every weekday morning.

Weekly digest contents

Configuring digest recipients and timing

6. Triage and prioritization

The PM Agent can triage incoming issues automatically based on rules you define. When new tickets arrive, it reads the title and description, assigns a priority, labels it, and routes it to the correct assignee or team — without you touching it.

Setting up auto-triage rules

AI-assisted triage

For tickets with enough context, the agent uses AI to assess severity more accurately than keyword matching. It considers the description, affected feature, user impact, and similar past incidents. You can always override its assessment: "Change ticket LIN-492 to P1 — the affected customer is enterprise tier."

7. Postmortems

After every SEV1 or SEV2 incident resolves, the PM Agent writes a draft postmortem within one hour. The draft includes a timeline reconstructed from the war room chat, the root cause analysis, customer impact assessment, and action items.

Postmortem structure

Reviewing and publishing the postmortem

The agent posts the draft to your session chat. Review it, suggest edits via chat ("Add that the root cause was the memory leak introduced in deploy d4a91f"), and when satisfied: "Publish the postmortem to our Notion page and email it to the team."

Action item tracking

Action items from postmortems are automatically created as tickets in your project board. The agent tracks them like any other ticket and alerts you if they go stale or approach their due date without progress.

8. Cost tracking

The PM Agent can track the engineering cost of incidents and unplanned work. This helps justify infrastructure investment and gives leadership visibility into how much reactive work is consuming sprint capacity.

What cost tracking calculates

Setting up cost tracking

9. Communicating with your agent

Your PM Agent session has a dedicated email address (pm-a3f9b2c1@agents.boboyka.com) and an in-session chat panel at hire.boboyka.com/session/[id]. Both share the same history.

Things you can ask the agent

10. Tier capabilities

11. Troubleshooting

The agent is not picking up new tickets from Linear

Linear's API requires the agent to poll periodically (webhooks not yet supported on all team plans). The default poll interval is 5 minutes on Pro, 15 minutes on Starter. If a ticket still does not appear after 15 minutes, tell the agent: "Refresh your Linear sync now — I just created issue LIN-500."

The agent declared the wrong severity

You can always override: "Change incident INC-2024-050 from SEV3 to SEV1 — this is affecting all paid customers." The agent will re-page appropriately and update the war room.

Digest is not being sent to Slack

Check that the Slack bot was invited to the digest channel. The agent needs to be a member of any channel it posts to. Tell the agent: "Diagnose why the digest is not posting to #engineering-weekly." It will report the exact Slack API error.

I want to exclude a ticket from SLA tracking

The postmortem is missing key information

The agent reconstructs the timeline from your incident channel messages and monitoring events. If critical events happened outside those channels (e.g., in a private DM or a phone call), it cannot know about them. Add them manually: "Add to the postmortem timeline at 14:32 UTC: on-call noticed the alert but was in a meeting for 10 minutes before responding."