ericrisco/rsc-harness / skills/monitoring
monitoring
VerifiedUse when setting up uptime and health monitoring, alerts, or on-call basics for a service already in production, so you learn it is down before customers do — health and readiness probes, alerting on SLO error-budget burn rather than raw counts, curing alert fatigue, on-call rotation with escalation, and a status page. NOT instrumenting logs, metrics or traces, or wiring OpenTelemetry (that is `observability`).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.0 KB |
| evals/README.md | markdown | 1021 B |
| evals/cases.yaml | config | 3.2 KB |
| references/burn-rate-and-oncall.md | markdown | 5.0 KB |
| references/tool-setup.md | markdown | 5.2 KB |
| scripts/verify.shexecutable | script | 5.7 KB |
scripts/verify.sh
View on GitHub#!/usr/bin/env bash
# verify.sh — monitoring config sanity gate. Read-only. Run inside YOUR project or
# point it at a monitoring config/spec dir or file. CI runs the same file (parity).
#
# Usage:
# bash scripts/verify.sh [PATH] # PATH defaults to "."
#
# What it checks, against the load-bearing rules of the monitoring skill:
# 1. Discover candidate monitoring files (alert rules, health specs, monitor configs).
# No candidates -> nothing to gate -> exit 0 (clean target is never a failure).
# 2. Health endpoints separate liveness from readiness (livez/liveness AND readyz/readiness).
# 3. At least one alert is symptom/SLO/burn-rate based with TWO windows — FAIL only if
# alert rules exist and the ONLY thresholds are bare cpu/memory.
# 4. Every alert block carries a runbook reference (runbook / runbook_url).
# 5. Banlist: Opsgenie introduced as a NEW setup; a monitor whose only target is "/".
#
# Dependency-light: bash + grep + find only. Portable to macOS bash 3.2 and CI bash 5.
# Read-only: never writes, never mutates. Exits nonzero only on a real FAIL.
#
# Env: NO_COLOR=1 to disable color.
set -euo pipefail
TARGET="${1:-.}"
if [[ -n "${NO_COLOR:-}" ]]; then YEL=""; GRN=""; RED=""; RST=""
else YEL=$'\033[33m'; GRN=$'\033[32m'; RED=$'\033[31m'; RST=$'\033[0m'; fi
ok() { printf '%sPASS:%s %s\n' "$GRN" "$RST" "$*"; }
fail() { printf '%sFAIL:%s %s\n' "$RED" "$RST" "$*"; }
skip() { printf '%sSKIP:%s %s\n' "$YEL" "$RST" "$*"; }
FAILED=0
# 1. discover candidate files. bash-3.2-safe (no mapfile); NUL-delimited for spaces.
FILES=()
if [[ -f "$TARGET" ]]; then
FILES+=("$TARGET")
elif [[ -d "$TARGET" ]]; then
while IFS= read -r -d '' f; do
FILES+=("$f")
done < <(find "$TARGET" \
\( -path '*/node_modules/*' -o -path '*/.git/*' -o -path '*/vendor/*' \) -prune \
-o -type f \( -name '*.yml' -o -name '*.yaml' -o -name '*.json' \
-o -name '*.rules' -o -name '*.conf' -o -name '*.md' \) -print0 2>/dev/null)
else
skip "target not found: $TARGET — nothing to gate"; exit 0
fi
# Keep only files that look like monitoring artifacts (alerts/health/monitors).
CAND=()
for f in ${FILES[@]+"${FILES[@]}"}; do
if grep -liE 'alert|burn[_-]?rate|livez|readyz|liveness|readiness|monitor|uptime|escalation|runbook' "$f" >/dev/null 2>&1; then
CAND+=("$f")
fi
done
if [[ ${#CAND[@]} -eq 0 ]]; then
printf 'No monitoring config/spec files found under %s — nothing to gate.\n' "$TARGET"
exit 0
fi
# Build a single haystack of the candidate contents for cross-file checks.
HAY="$(cat ${CAND[@]+"${CAND[@]}"} 2>/dev/null || true)"
# Helpers (case-insensitive grep against the haystack).
has() { printf '%s' "$HAY" | grep -qiE "$1"; }
# 2. liveness vs readiness — only meaningful if a health spec is present at all.
if printf '%s' "$HAY" | grep -qiE 'livez|liveness|readyz|readiness|health'; then
if has 'livez|liveness' && has 'readyz|readiness'; then
ok "health spec separates liveness from readiness"
else
fail "health spec does not separate liveness (/livez) from readiness (/readyz)"
FAILED=1
fi
else
skip "no health-endpoint spec present — liveness/readiness check not applicable"
fi
# 3. symptom/SLO/burn-rate alert with two windows.
if has 'alert|expr|burn|rate\(|error[_ ]?rate|severity'; then
TWO_WINDOW=0
# burn-rate keyword, or an explicit long+short pairing of distinct time windows.
if has 'burn[_-]?rate|multi[_-]?window'; then TWO_WINDOW=1; fi
if printf '%s' "$HAY" | grep -qE '\[(1h|6h|3d|24h)\]' \
&& printf '%s' "$HAY" | grep -qE '\[(5m|30m|6h)\]'; then TWO_WINDOW=1; fi
if has '(long|short)[ _-]?window'; then TWO_WINDOW=1; fi
# Are the ONLY thresholds bare cpu/memory? (cause-based, no symptom alert).
ONLY_CAUSE=0
if has 'cpu|memory|mem_usage|load average' \
&& ! has 'error[_ ]?rate|5\.\.|5xx|latency|p9[59]|burn|availability|uptime'; then
ONLY_CAUSE=1
fi
if [[ $ONLY_CAUSE -eq 1 ]]; then
fail "only cause-based thresholds found (cpu/memory) — add a symptom/SLO alert"
FAILED=1
elif [[ $TWO_WINDOW -eq 1 ]]; then
ok "symptom/SLO alert with two windows (burn-rate or long+short) present"
else
fail "alert present but no two-window/burn-rate symptom alert — single-window 'by vibes'"
FAILED=1
fi
# 4. runbook reference on alerts.
if has 'runbook'; then
ok "alerts reference a runbook"
else
fail "alert rules carry no runbook/runbook_url reference"
FAILED=1
fi
else
skip "no alert rules present — alert/runbook checks not applicable"
fi
# 5a. banlist — Opsgenie as a NEW setup (EOL 2027-04-05).
# Allow mentions that are clearly a do-NOT note; flag a bare adoption.
if has 'opsgenie'; then
if printf '%s' "$HAY" | grep -iE 'opsgenie' | grep -qiE "do not|don't|deprecat|eol|retir|avoid|instead of|not start|no usar|migrat"; then
ok "opsgenie referenced only as a do-not/deprecation note"
else
fail "opsgenie configured as a new setup — it is EOL 2027-04-05; use PagerDuty/incident.io/Better Stack/Grafana Cloud IRM"
FAILED=1
fi
fi
# 5b. banlist — a monitor whose only target is the homepage "/".
# Flag url:"/" or path "/" used as a probe target with no critical-journey/readyz target.
if printf '%s' "$HAY" | grep -qE '(url|path|target|target_url)["'"'"' :]+["'"'"']?/["'"'"']?\s*$' \
|| printf '%s' "$HAY" | grep -qE '(url|path|target)["'"'"' :]+["'"'"']/["'"'"']'; then
if has 'readyz|/api/|journey|checkout|login|critical'; then
ok "homepage probe present but a deeper/critical-journey target also configured"
else
fail "monitor target appears to be homepage '/' only — probe the critical user journey, not just /"
FAILED=1
fi
fi
# summary
if [[ $FAILED -eq 0 ]]; then
ok "all monitoring checks passed (skips are not failures)"
else
fail "one or more monitoring checks failed"
fi
exit "$FAILED"
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |