Skill
core:ship-and-monitor
Description
A skill that commits changes, pushes, opens a PR, and then uses /loop to monitor CI/CD at 1-minute intervals — auto-fixing obvious failures and stopping to ask the user for non-obvious ones.
Proposed Behavior
Workflow
- Confirm branch — Before committing, verify the current branch is correct. In long conversations that span multiple branches, never add commits to a branch that was already merged. Do a quick
git status / git log sanity check and confirm with the user if anything looks off.
- Commit and push — Stage relevant files (never
git add -A), create a commit, push with -u.
- Open PR — Target
develop by default (or user-specified branch) via gh pr create. Show the PR URL.
- Start
/loop 1m — Monitor CI/CD checks using gh pr checks. On pass, report success and stop. On failure, diagnose and either fix or ask.
Auto-fix (obvious failures)
- TypeScript/build errors (type mismatches, missing imports, unused vars)
- Lint/format violations
- Test failures clearly caused by our changes (wrong assertion, missing mock update)
- Missing documented environment variables
Stop and ask (non-obvious failures)
- Test failures in code we didn't touch
- Infrastructure/deployment failures (Docker, Terraform, Azure)
- Flaky or intermittent test failures
- Permission/auth errors
- Fixes that would require architectural changes or change behavior beyond original intent
Branch safety
- Always verify current branch before committing
- In long conversations, confirm the user hasn't moved on from a previously-merged branch
- Never blindly commit to whatever branch happens to be checked out — validate it's the intended target
Examples
User says: "ship it"
Skill: confirms branch → commits → pushes → opens PR → starts /loop 1m monitoring
User says: "commit push and watch CI"
Skill: same flow
CI lint fails:
Skill: reads log, fixes lint error, commits, pushes, continues monitoring
CI deploy fails:
Skill: stops loop, reports failure details, asks user for guidance
Skill
core:ship-and-monitorDescription
A skill that commits changes, pushes, opens a PR, and then uses
/loopto monitor CI/CD at 1-minute intervals — auto-fixing obvious failures and stopping to ask the user for non-obvious ones.Proposed Behavior
Workflow
git status/git logsanity check and confirm with the user if anything looks off.git add -A), create a commit, push with-u.developby default (or user-specified branch) viagh pr create. Show the PR URL./loop 1m— Monitor CI/CD checks usinggh pr checks. On pass, report success and stop. On failure, diagnose and either fix or ask.Auto-fix (obvious failures)
Stop and ask (non-obvious failures)
Branch safety
Examples
User says: "ship it"
Skill: confirms branch → commits → pushes → opens PR → starts
/loop 1mmonitoringUser says: "commit push and watch CI"
Skill: same flow
CI lint fails:
Skill: reads log, fixes lint error, commits, pushes, continues monitoring
CI deploy fails:
Skill: stops loop, reports failure details, asks user for guidance