Analyzes cloud usage across Neon, Vercel, Railway, and GitHub Actions to produce ranked cost recommendations.
npx skills add https://github.com/aaronjmars/aeon --skill spend-watch
Please help me install this Agent Skill. Skill: spend-watch Source: https://github.com/aaronjmars/aeon/tree/main/skills/spend-watch Command: npx skills add https://github.com/aaronjmars/aeon --skill spend-watch
Cloud bills across multiple platforms are hard to attribute and act on: usage data is scattered across Neon, Vercel, Railway, and GitHub Actions, and most tools invent dollar figures where no billing API exists. This Skill pulls per-object usage from each platform, attributes spend to the biggest drivers, root-causes each one, and emits recommendations ranked by real signal (idle %, over-allowance, failure rate) with dollar figures only where the billing API returns real ones.
Ask the agent to run spend-watch with scope 'all' to audit every configured platform and receive a ranked cost recommendation digest.
High-intent search queries and answers about installing and using this skill.
Run spend-watch with an empty or 'all' scope to sweep every platform whose API token is present. It pulls per-object usage from each provider, attributes spend to the biggest drivers, and emits one combined digest of ranked recommendations.
The actions adapter checks your account-wide minutes against the included allowance, attributes usage per repo and workflow, and recommends trims like reducing cron frequency or moving macOS jobs to Ubuntu. Armed mode can delete stale artifacts or open a PR trimming the biggest schedule.
No, by default it is read-only and only recommends. Mutations require the arm: prefix (for example arm:neon), and even then each adapter re-reads the target's live state before acting and never deletes production deployments or default branches.
Neon's lower plans and Vercel's public API expose no billing dollars, so the Skill reports real signals instead, such as idle-awake percentage or stale-preview count. It never fabricates a dollar figure by multiplying usage by an assumed rate.
It uses NEON_API_KEY, VERCEL_TOKEN, and RAILWAY_TOKEN through secretcurl placeholders, plus GH_TOKEN for the GitHub Actions adapter via the gh CLI. Any adapter whose secret is absent skips itself and logs the skip without failing the run.
Yes, append dry-run to the scope variable to build the full digest while suppressing the notify step. This is intended for testing the analysis output before enabling real notifications.
The-Entourage-AI-DevelopmentProvision deployment environments for Entourage projects across Vercel, Railway, Neon, Clerk, and DNS.
dull-birdFetch RSS feeds, deduplicate titles, rank AI news, and write Markdown digests to Obsidian.
BagelHoleDefine AI incident classes, severity frameworks, and response playbooks for LLM outages.
ramp-publicMerge card transactions and bill payments into vendor spend totals.
