lifestep.io · service status
Checks and small services built while running Claude Code and Codex workers in parallel — plus four open APIs with no signup, no API key and no tracking. This page checks the APIs live, every time you load it.
Paste a CLAUDE.md or AGENTS.md and see which rules an agent
cannot act on: truncation risk, headings with no rule under them, instructions with no
verification command, prohibitions with no stated alternative. Static checks, in your
browser — nothing is uploaded. Checks it cannot evaluate say not evaluated, never
pass.
Tells DONE, FAILED and STALL apart when a worker
exits. A worker that stopped to ask for approval exits cleanly and reads like success; this
checks the recorded exit code and a completion marker.
brew tap soul-sol/tap
brew install agent-watchTen adversarial review prompts, twelve CLAUDE.md operating rules and a task
brief, inserted at your cursor from the Command Palette. It inserts text; it does not run an
agent or send your code anywhere.
Two GitHub Actions from the same practice: a completion gate that refuses a run without
evidence, and a binary-architecture check that catches exec format error before
the binary is ever run.
Reads an .xlsx and returns sheet names, dimensions, formulas, defined names, and external links — without Excel and without uploading to a third party.
Syntax, MX records, disposable-domain detection, and typo suggestions for common providers. Validation only — it never sends mail to the address.
One GET returns an SVG or PNG QR code. Size, margin, and error
correction are query parameters, so it drops straight into an <img> tag.
A, MX, NS and TXT records, SPF and DMARC policy parsing, and TLS certificate expiry for any domain — one request instead of four dig invocations.
These services are built and operated by AI coding agents running in parallel. When one of them reports success while the system is in a different state, the incident gets written up with the exact check that catches it next time: Seven ways my coding agents reported success while failing.
The same practice these services were built under — briefing coding agents, catching the ones that stall, and reviewing what they write — is written up as a set of manuals, review prompts and templates: the Complete Agent Ops Kit ($29), or the individual titles at the LifeStep store. The four APIs above stay free either way.