ed-ai-teammate

name	ed-ai-teammate
description	AI Teammate - manage connectors (PagerDuty, Slack, GitHub, ...), update teammates (agents) and view teammate activity.
metadata	{"version":"1.0.0","author":"edgedelta","repository":"https://github.com/edgedelta/agent-skills","tags":"edgedelta,ai,teammate,connectors","alwaysApply":"false"}

Edge Delta AI Teammate

The AI Teammate ingests signals from connected tools (PagerDuty, Slack, GitHub, ...) and acts on them. This skill manages those connectors and inspects teammate activity.

Prerequisites

The edx CLI must be installed and authenticated. See the ed-edx skill.

Inspect

edx ai connectors list             # configured connectors
edx ai connectors specs            # available connector types + required fields
edx ai connectors environments     # where connectors can run
edx ai activity --lookback 24h     # teammate activity metrics
edx ai agents list                 # AI Teammates (agents); alias: edx ai teammates
edx ai agents get <agent-id>       # a single teammate's full definition

Update a Teammate (Agent)

edx ai agents update prompts for confirmation unless --yes.

Just the prompts (the common case)

Use the --*-prompt flags — each takes an inline string or @file (@- for stdin). The command reads the current teammate, backfills whichever prompt you did not pass (the service requires both masterPrompt and userPrompt on every update), and sends only the prompts. Model, temperature, tools and everything else are left untouched, so you never deal with model-tuning validation:

edx ai agents update <id> --master-prompt @master.md --user-prompt @user.md
edx ai agents update <id> --master-prompt "You are a concise SRE assistant."

Any field (clone-and-edit)

Like edx monitors update: fetch, edit the JSON, apply it back.

edx ai agents get <agent-id> > agent.json      # edit fields under "data":
                                               # model, toolConfigurations,
                                               # connectors, priority, ...
edx ai agents update <agent-id> --file agent.json

The whole get envelope is accepted and unwrapped automatically. Only the fields you include are changed; a field set to null is cleared. Use -f/--file - to read from stdin. Note: re-sending the full object re-validates model-tuning fields (model, modelTemperature) even if untouched — to change only prompts prefer the flags above, or send just the prompt fields:

edx ai agents get <id> | jq '.data | {masterPrompt,userPrompt,toolingPrompt}' \
  | edx ai agents update <id> --file - --yes

Add or Update a Connector

1. Find the connector type and its required fields:

edx ai connectors specs --output json | jq '.[] | select(.type=="pagerduty")'

2. Build the request JSON per the spec (type, name, credentials/settings).

3. Apply it:

edx ai connectors update --file connector.json

Connector data flows through an ingestion pipeline that Edge Delta provisions automatically; check it with edx pipelines list --keyword ai.

Remove a Connector

The delete request body identifies the connector (same shape as update):

edx ai connectors delete --file connector.json --yes

Troubleshooting

Problem	Fix
Connector not ingesting	edx pipelines list --keyword ai then edx health problems
Unknown required fields	edx ai connectors specs is the source of truth
Credential errors	Re-apply with edx ai connectors update --file and fresh secrets
Teammate update rejected (4xx)	Re-fetch with edx ai agents get <id>, edit only the data fields, re-apply