semaphore-debugging

name: semaphore-debugging description: Debug and triage Semaphore CI pipelines/jobs with sem-ai, falling back to the legacy sem CLI only for interactive attach/debug flows that sem-ai does not cover.

Semaphore Debugging

Use this skill for CI troubleshooting, stuck pipelines, flaky jobs, log inspection, and safe recovery actions.

Prefer sem-ai for structured inspection and mutation. Use the legacy sem CLI only when there is no equivalent sem-ai command, mainly interactive job shells.

Trigger

• CI is failing, flaky, stuck, or slow.
• The request mentions Semaphore, CI status, pipeline/workflow/job ids, logs, test failures, reruns, stops, promotions, or deployment state.
• You need real Semaphore runtime state before taking action.

Session Start

sem-ai connect simplepractice.semaphoreci.com "$SEMAPHORE_API_TOKEN"
sem-ai context show

Fallback if env var is missing:

sem-ai connect simplepractice.semaphoreci.com <YOUR_API_TOKEN>

sem-ai shares the same ~/.sem.yaml context file as the legacy sem CLI.

Fast Incident Workflow

# Let sem-ai collect the first pass of evidence.
sem-ai diagnose <workflow-id>

# If the workflow id is not known yet:
sem-ai status --project simplepractice --branch <branch>
sem-ai workflow list --project simplepractice --branch <branch>

# Drill down when diagnose is not enough.
sem-ai workflow show <workflow-id>
sem-ai pipeline show <pipeline-id>
sem-ai job log <job-id>
sem-ai test summary --pipeline <pipeline-id>

# Recovery after inspection.
sem-ai pipeline rebuild <pipeline-id>
sem-ai workflow rerun <workflow-id>
sem-ai watch <workflow-id>

Command Quick Reference

# Discovery
sem-ai status --project simplepractice --branch <branch>
sem-ai workflow list --project simplepractice
sem-ai pipeline list --project simplepractice
sem-ai job list

# Inspection
sem-ai workflow show <workflow-id>
sem-ai pipeline show <pipeline-id>
sem-ai pipeline topology <pipeline-id>
sem-ai job show <job-id>
sem-ai job log <job-id>
sem-ai test summary --pipeline <pipeline-id>
sem-ai test report --pipeline <pipeline-id>
sem-ai blast-radius <pipeline-id>
sem-ai critical-path <pipeline-id>

# Recovery
sem-ai job stop <job-id>
sem-ai pipeline stop <pipeline-id>
sem-ai pipeline rebuild <pipeline-id>
sem-ai workflow rerun <workflow-id>
sem-ai rerun-failed <pipeline-id>
sem-ai watch <workflow-id>

Legacy sem Fallback

Use sem only when the needed workflow is not available in sem-ai.

# Running jobs only.
sem attach <job-id>

# Finished jobs only; then source the reproduced command environment.
sem debug <job-id>
source ~/commands.sh

Do not use sem get or sem logs by default anymore. The equivalent inspection path is sem-ai workflow show, sem-ai pipeline show, and sem-ai job log.

Guardrails

• Read before mutate: inspect with sem-ai diagnose, sem-ai workflow show, sem-ai pipeline show, or sem-ai job log before stop/rerun/rebuild.
• sem-ai defaults to JSON. Use --format table for human scanning and JSON for scripts.
• Promotion commands require --confirm; do not add it until the target, parameters, and pipeline id have been verified.
• If command shape is unclear, run sem-ai <command> --help or sem-ai discover before guessing.

Repo-Aware Check

CI behavior is defined by repo files. Confirm before assumptions:

• <app-repo>/.semaphore/ci.yml
• <app-repo>/.semaphore/*.sh
• <app-repo>/scripts/*ci*

For simplepractice, treat change_in and branch-specific rules as the source of truth from CI config.

Reference: https://docs.semaphore.io/reference/sem-ai-cli