Add /debug-ci and /pipeline-status skills

.claude/skills/debug-ci/skill.md

@@ -0,0 +1,200 @@
+---
+name: debug-ci
+description: Diagnose CI pipeline failures and suggest fixes
+metadata:
+  argument-hint: "[pr-number|run-id|job-name]"
+---
+
+# Debug CI
+
+Diagnose CI pipeline failures by fetching logs from GitHub Actions, identifying the root cause, and suggesting local reproduction steps.
+
+## Usage
+
+```
+/debug-ci              # Diagnose the latest CI failure on the current branch
+/debug-ci 3218         # Diagnose CI for a specific PR
+/debug-ci clippy       # Diagnose a specific job type
+```
+
+## CI Jobs Reference
+
+The `ci.yml` workflow runs these jobs (all depend on `set-variables`):
+
+| Job | Command | Runner | Blocks merge | Notes |
+|-----|---------|--------|--------------|-------|
+| `fmt` | `cargo +nightly fmt --all -- --check` | ubuntu-latest | Yes | |
+| `clippy` | `SKIP_WASM_BUILD=1 cargo clippy --all-targets --locked --workspace` | ubuntu-latest | No (`continue-on-error`) | Rust cache disabled |
+| `spellcheck` | `cargo spellcheck check` | ubuntu-latest | Yes | Excludes `codegen_runtime.rs`, `weights.rs` |
+| `check` | `SKIP_WASM_BUILD=1 cargo check --locked --workspace` | ubuntu-latest | Yes | Rust cache disabled |
+| `test` | `cargo test --workspace` | parity-large | Yes | Uses Rust cache |
+| `deny` | `cargo deny check advisories bans sources` | ubuntu-latest | No (`continue-on-error`) | |
+| `deny-licenses` | `cargo deny check licenses` | ubuntu-latest | Yes | |
+| `check-rustdocs` | `cargo doc --no-deps --all --workspace --document-private-items` | ubuntu-latest | Yes | Uses Rust cache |
+| `build` | `cargo build --release --workspace` | parity-large | Yes | Uses Rust cache |
+| `build_docker` | Docker build (no push) | ubuntu-latest | Yes | Depends on `build` |
+
+## Procedure
+
+### Step 1: Identify the CI Run
+
+Determine which run to debug based on `$ARGUMENTS`:
+
+**If a PR number** (all digits, typically 4 digits):
+
+```bash
+gh run list --workflow=ci.yml --branch=<PR_HEAD_BRANCH> --limit=3
+```
+
+Or look up the PR first:
+
+```bash
+gh pr view <NUMBER> --json headRefName --jq '.headRefName'
+```
+
+**If a run ID** (large number):
+
+```bash
+gh run view <RUN_ID>
+```
+
+**If a job name** (e.g. `clippy`, `test`, `fmt`):
+Find the latest run on the current branch, then focus on that specific job.
+
+**If no arguments**: use the current branch:
+
+```bash
+gh run list --workflow=ci.yml --branch=$(git branch --show-current) --limit=3
+```
+
+If the current branch is `master`, show the latest failed run across all branches:
+
+```bash
+gh run list --workflow=ci.yml --status=failure --limit=5
+```
+
+Pick the most recent failed (or in-progress) run and note its `<RUN_ID>`.
+
+### Step 2: Get Run Overview
+
+Fetch the run summary to see which jobs passed and which failed:
+
+```bash
+gh run view <RUN_ID>
+```
+
+List all jobs and their statuses. Identify the failed job(s).
+
+If all jobs passed, report success and stop.
+
+### Step 3: Fetch Failed Job Logs
+
+For each failed job, fetch the logs:
+
+```bash
+gh run view <RUN_ID> --log-failed
+```
+
+If the output is too large, target a specific job:
+
+```bash
+gh run view <RUN_ID> --job=<JOB_ID> --log
+```
+
+You can find job IDs from:
+
+```bash
+gh api repos/{owner}/{repo}/actions/runs/<RUN_ID>/jobs --jq '.jobs[] | "\(.id) \(.name) \(.conclusion)"'
+```
+
+### Step 4: Diagnose the Failure
+
+Analyze the error output and classify it:
+
+**Formatting (`fmt`)**:
+- Error: "Diff in ..." — code is not formatted
+- Fix: `cargo +nightly fmt --all`
+
+**Linting (`clippy`)**:
+- Parse clippy warnings/errors from the output
+- Fix: `SKIP_WASM_BUILD=1 cargo clippy --all-targets --locked --workspace --fix --allow-dirty`
+- Note: this job has `continue-on-error: true`, so it doesn't block merge
+
+**Spellcheck**:
+- Error: misspelled words in `.rs` files
+- Fix: add words to `.config/spellcheck.toml` or fix the spelling
+- Note: excludes `codegen_runtime.rs` and `weights.rs`
+
+**Compile check (`check`)**:
+- Error: compilation errors
+- Fix: `SKIP_WASM_BUILD=1 cargo check --locked --workspace`
+- Common causes: missing imports after codegen, type mismatches after runtime upgrade
+
+**Tests (`test`)**:
+- Parse test failure output: look for `FAILED`, `panicked at`, `test result: FAILED`
+- Fix: `cargo test --workspace` locally
+- Note: runs on parity-large runner, so resource-dependent tests may behave differently locally
+
+**Dependency audit (`deny`)**:
+- Error: advisory/ban/source violations
+- Fix: `cargo deny check advisories bans sources`
+- Note: this job has `continue-on-error: true`, so it doesn't block merge
+
+**License check (`deny-licenses`)**:
+- Error: license violation
+- Fix: `cargo deny check licenses` — update `deny.toml` exceptions or change the dependency
+- This one IS strict (blocks merge)
+
+**Rustdocs (`check-rustdocs`)**:
+- Error: documentation build failures (broken links, missing docs)
+- Fix: `cargo doc --no-deps --all --workspace --document-private-items`
+
+**Build**:
+- Error: release build failure
+- Fix: `cargo build --release --workspace`
+- Common causes: same as `check`, but release-only optimizations can surface additional issues
+
+**Docker build (`build_docker`)**:
+- Error: Dockerfile issues, missing artifacts
+- Fix: check `ci.Dockerfile` and ensure `build` job artifacts are correct
+
+### Step 5: Check for Known Issues
+
+Look for patterns that indicate systemic issues rather than code problems:
+
+- **Rust cache disabled**: clippy and check jobs have cache disabled (`# todo: fixme`). If these jobs are slow or timing out, this is why.
+- **Runner availability**: `parity-large` runner is needed for test and build. If these jobs are queued for a long time, it may be a runner capacity issue.
+- **Concurrency cancellation**: the workflow cancels in-progress runs when a new push arrives. If a run shows as "cancelled", this is expected.
+- **Fork PRs**: env variables don't work for fork PRs (handled via `set-variables` job outputs). If the CI image isn't resolving, check if the PR is from a fork.
+
+### Step 6: Output Report
+
+Present a structured report:
+
+```
+## CI Diagnosis: <branch or PR>
+
+**Run**: <RUN_ID> (<status>)
+**Triggered by**: <push to master | PR #N>
+
+| Job | Status | Duration |
+|-----|--------|----------|
+| fmt | pass/fail | Xs |
+| clippy | pass/fail (soft) | Xs |
+| ... | ... | ... |
+
+### Failure: <job name>
+
+**Error**:
+<key error lines from the log>
+
+**Root cause**: <analysis>
+
+**Fix**:
+<specific commands to run locally>
+
+### Recommendations
+- <actionable next steps>
+```
+
+If the failure is in a `continue-on-error` job (clippy, deny), note that it doesn't block merge but should still be fixed.

.claude/skills/pipeline-status/skill.md

@@ -0,0 +1,134 @@
+---
+name: pipeline-status
+description: Show CI/CD pipeline health at a glance
+metadata:
+  argument-hint: ""
+---
+
+# Pipeline Status
+
+Show the overall health of all CI/CD pipelines, the currently deployed version, and any pending actions.
+
+## Usage
+
+```
+/pipeline-status
+```
+
+No arguments required. Shows a full overview of all workflows, deployment state, and release readiness.
+
+## Procedure
+
+### Step 1: Gather Workflow Status
+
+Fetch recent runs for all three workflows in parallel:
+
+```bash
+gh run list --workflow=ci.yml --limit=5
+```
+
+```bash
+gh run list --workflow=build-tag.yml --limit=5
+```
+
+```bash
+gh run list --workflow=deploy.yml --limit=5
+```
+
+For each workflow, note:
+- Last run status (success, failure, in_progress, cancelled)
+- When it ran
+- What triggered it (branch, tag, manual)
+
+### Step 2: Identify Current Deployed Version
+
+Find the last successful deploy:
+
+```bash
+gh run list --workflow=deploy.yml --status=success --limit=1
+```
+
+Extract which version tag was used from the run's head branch/tag.
+
+### Step 3: Identify Latest Release
+
+Check the latest git tag and GitHub Release:
+
+```bash
+git tag -l 'v*' --sort=-v:refname | head -3
+```
+
+```bash
+gh release list --limit=3
+```
+
+Compare: is the latest release deployed? If the latest tag is newer than the last deploy, flag it.
+
+### Step 4: Check Current Branch CI
+
+Show CI status for the current branch (if not master):
+
+```bash
+gh run list --workflow=ci.yml --branch=$(git branch --show-current) --limit=3
+```
+
+### Step 5: Check Docker Image Availability
+
+Verify Docker images exist for the latest tag:
+
+```bash
+curl --fail --silent "https://hub.docker.com/v2/repositories/paritytech/substrate-relay/tags/$(git tag -l 'v*' --sort=-v:refname | head -1)" | head -c 200
+```
+
+### Step 6: Read Relay Version from Source
+
+```bash
+grep '^version' substrate-relay/Cargo.toml
+```
+
+Compare the Cargo.toml version against the latest git tag. If the Cargo.toml version is ahead, a release may be in progress.
+
+### Step 7: Output Dashboard
+
+Present a structured dashboard:
+
+```
+## Pipeline Status
+
+### Workflows
+
+| Workflow | Last Run | Status | Trigger | When |
+|----------|----------|--------|---------|------|
+| CI | #<id> | pass/fail | PR #N / push to master | <time> |
+| Build Tag | #<id> | pass/fail | tag v<X> | <time> |
+| Deploy | #<id> | pass/fail | manual (v<X>) | <time> |
+
+### Versions
+
+| | Version |
+|---|---------|
+| Source (Cargo.toml) | <version> |
+| Latest tag | v<version> |
+| Latest release | v<version> |
+| Deployed | v<version> |
+| Docker image | exists/missing |
+
+### Current Branch: <branch>
+
+CI status: <pass/fail/pending/none>
+
+### Action Items
+
+- <any issues that need attention>
+```
+
+Possible action items to flag:
+- **CI failing on master**: "master CI is red — run `/debug-ci` to investigate"
+- **Unreleased version bump**: "Cargo.toml (v1.8.16) is ahead of latest tag (v1.8.15) — run `/release-finalize` if PR is merged"
+- **Undeployed release**: "Latest tag v1.8.16 is not deployed — run `/deploy v1.8.16`"
+- **Missing Docker image**: "Docker image for v<X> not found — check `build-tag.yml` run"
+- **Build-tag failure**: "Last `build-tag.yml` run failed — Docker image may not have been pushed"
+- **Stale deploy**: "Last deploy was >30 days ago — verify relay health with `/check-health all`"
+- **Branch CI red**: "Current branch CI is failing — run `/debug-ci` to investigate"
+
+If everything looks healthy, report: "All pipelines green. No action needed."