Menu
Simulate fulmine GitHub Actions CI pipeline locally. Runs lint, vet, build, unit tests, and integration tests (MANDATORY). Optionally proto lint and DB vet — matching the real CI checks that run on PRs.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Analyze a project repository and add it to the Arkadian documentation registry with standardized structure. Use when: user provides a path to a new project to onboard.
Analyse an Arkadian orchestrator session - read transcripts, logs, and artifacts to understand what happened and suggest improvements. Use when: user wants to debug or review a past session.
Generate validated operational SOPs for Ark ecosystem projects by testing commands before documenting them. Use when: user wants to create or audit SOPs for a project.
Remove a project from the Arkadian documentation registry and delete all associated documentation files. Use when: user wants to deregister a project.
Update project documentation based on new commits and changes in the repository. Use when: user wants to sync docs after project changes.
Self-improvement loop for arkadian. Launches arkadian -d with a GitHub issue, polls until done, evaluates per-agent criteria, reads transcripts on failure, improves prompts, deletes worktree, and relaunches. Loops until passing or max iterations.
| name | fulmine-gha |
| description | Simulate fulmine GitHub Actions CI pipeline locally. Runs lint, vet, build, unit tests, and integration tests (MANDATORY). Optionally proto lint and DB vet — matching the real CI checks that run on PRs. |
Run the same checks that fulmine's GitHub Actions CI pipeline runs on pull requests. This catches CI failures locally before pushing.
Source of truth: ${FULMINE_REPO}/.github/workflows/ — always check these files for the latest CI definition. This skill documents the known pipeline as a fallback.
Before running CI checks, verify you're in the right location:
# Must run from the worktree directory (inherited from implement phase)
pwd
ls go.mod # Must exist — confirms we're in a fulmine repo root
make lint
Uses golangci-lint. Auto-fixes formatting issues.
On failure:
make lint again — it auto-fixes some issuesmake lint to confirmmake vet
Static analysis across all packages.
make build-static-assets
Builds templates and web assets. Must succeed before binary build.
make build
Compiles the fulmine binary. Must succeed.
make test
Runs all unit tests with race detector, excluding internal/test/e2e. Safe to use for fulmine (no port conflict unlike arkd).
Only run if .proto files were modified:
git diff --name-only HEAD~1 | grep '\.proto$' && PROTO_CHANGED=true || PROTO_CHANGED=false
if [ "$PROTO_CHANGED" = "true" ]; then
make proto-lint
fi
Only run if migration files were modified:
git diff --name-only HEAD~1 | grep 'migration' && MIG_CHANGED=true || MIG_CHANGED=false
if [ "$MIG_CHANGED" = "true" ]; then
make vet_db
fi
Reuse the infrastructure already running from the implement phase. Do NOT start new Docker containers.
# Verify infra is still running
curl -s http://localhost:7001/api/v1/info > /dev/null && echo "fulmine OK" || echo "fulmine NOT RUNNING"
curl -s http://localhost:7070/v1/info > /dev/null && echo "arkd OK" || echo "arkd NOT RUNNING"
curl -s http://localhost:3000/api/blocks/tip/height > /dev/null && echo "Esplora OK" || echo "Esplora NOT RUNNING"
If infrastructure is running, run a subset of integration tests as regression check:
# Uncomment TestMain if it was commented out during implement phase
# Then run a few key tests
go test -v -count=1 -run "TestVHTLC|TestDelegate" \
-timeout 600s -race -p=1 ./internal/test/e2e/...
If infrastructure is NOT running, you MUST start it before proceeding:
# Start the test environment
cd ${FULMINE_REPO} # Use the main repo, not the worktree
make build-test-env && make setup-test-env
cd - # Return to worktree
Wait for services to be ready (re-check health endpoints above). If infrastructure cannot be started after 3 attempts, report the CI phase as FAILED — do NOT skip integration tests. Integration tests are the most critical CI check.
Write ci-evidence.md to ${ARTIFACTS_DIR} with results of each check:
# CI Evidence
## Environment
- **Project:** fulmine
- **Branch:** <branch name>
- **Worktree:** <worktree path>
## Check Results
| Check | Status | Duration | Notes |
|-------|--------|----------|-------|
| Lint | PASS/FAIL | Xs | |
| Vet | PASS/FAIL | Xs | |
| Build Assets | PASS/FAIL | Xs | |
| Build | PASS/FAIL | Xs | |
| Unit Tests | PASS/FAIL | Xs | N passed, M failed |
| Proto Lint | PASS/SKIP | Xs | |
| DB Vet | PASS/SKIP | Xs | |
| Integration | PASS/FAIL | Xs | MANDATORY — never skip |
## Detailed Output
### Lint
<raw output>
### Vet
<raw output>
### Build Static Assets
<raw output>
### Build
<raw output>
### Unit Tests
<raw output or summary>
### Proto Lint
<raw output or "SKIPPED: no proto changes">
### DB Migration Vet
<raw output or "SKIPPED: no migration changes">
### Integration Tests
<raw output — MUST NOT be skipped>
## Auto-Fixes Applied
- <list of auto-fixes, or "None">
## Verdict
All checks: PASS / X of Y FAILED
{
"agent_specific": {
"lint_passed": true,
"vet_passed": true,
"build_assets_passed": true,
"build_passed": true,
"unit_tests_passed": true,
"proto_lint_passed": true,
"proto_lint_skipped": false,
"db_vet_passed": true,
"db_vet_skipped": true,
"integration_tests_passed": true,
"integration_tests_skipped": false,
"checks_run": 8,
"checks_passed": 8,
"checks_failed": 0,
"auto_fixes_applied": 0,
"files_changed": []
}
}