edpa-sync-people

name	edpa-sync-people
user-invocable	true
description	Reconcile .edpa/config/people.yaml against the repository's GitHub collaborator list. Reports adds/removes; on apply, flips removed collaborators to availability=unavailable and (with --auto-add) appends auto-filled stub entries for new ones via PR review.
license	MIT
compatibility	GitHub CLI (gh), Python 3.10+
allowed-tools	Read Bash(python3 *) Bash(gh *) Bash(git *)
metadata	{"author":"Jaroslav Urbánek","version":"1.0.0","domain":"governance","phase":"maintenance","standard":"AgentSkills v1.0"}

EDPA Sync People — Reconcile collaborators with people.yaml

What this does

Pulls the current GitHub collaborator list for the repo configured in .edpa/config/edpa.yaml (sync.github_org / sync.github_repo) and diffs it against .edpa/config/people.yaml:

• Adds: GitHub collaborator with no matching github field in people.yaml. By default the maintainer reviews them; with apply --auto-add the workflow appends an auto-filled stub (login, public name, public email) and leaves role/team/FTE blank.
• Removes: person in people.yaml whose github login is no longer on the repo. Their availability flips to unavailable (factual, no review needed).
• Unchanged: matched both sides — nothing to do.

Arguments

$ARGUMENTS = optional mode override:

• empty / status — read-only diff, no writes
• apply — flip availability for removes; leave adds for review
• apply --auto-add — also append stubs for adds (PR-friendly)

Steps

1. Verify prerequisites

gh auth status
test -f .edpa/config/people.yaml
test -f .edpa/config/edpa.yaml

2. Run the script

For status (default):

python3 .claude/edpa/scripts/sync_collaborators.py status

For apply (removes only — safe default):

python3 .claude/edpa/scripts/sync_collaborators.py apply

For apply with auto-add (appends stubs):

python3 .claude/edpa/scripts/sync_collaborators.py apply --auto-add

3. Review pending adds

When the script reports Adds pending review, each new collaborator needs role/team/FTE filled in by the maintainer before EDPA can credit them capacity. The auto-fill includes:

• id — derived from login (lowercase)
• name — from gh api users/{login} (or login fallback)
• email — from public profile (often blank)
• github — the login
• availability: confirmed

The maintainer fills in:

• role — Arch / Dev / DevSecOps / PM / QA
• team
• fte (e.g. 0.5, 1.0)
• capacity_per_iteration (FTE × hours_per_week × iteration_weeks)

4. Validate the result

python3 .claude/edpa/scripts/validate_iterations.py

validate_iterations.py runs both iteration AND people validators — warnings will surface anyone without a github login who is referenced as an assignee, anyone in people.yaml who is referenced nowhere, and any drift in iteration dates.

Automation

The same logic is wired up as a workflow that runs on GitHub member events:

• member.added / member.edited → workflow opens a PR with stubs pre-filled (you fill role/team/FTE and merge).
• member.removed → workflow commits the availability=unavailable flip directly on main (factual, no review needed).

See plugin/edpa/workflows/collaborators-sync.yml. The skill is the manual-trigger counterpart for back-fills, repairs, or initial sync on a brand-new EDPA install.

Error handling

• gh auth status failing → instruct: gh auth login
• Empty sync.github_org / github_repo in edpa.yaml → instruct user to run /edpa:setup first or pass --repo owner/repo explicitly
• Network / rate limit → the script returns exit 1 with stderr; re-running after a minute usually works