Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add jcwleo/oh-no-harness --skill ultraworkgit clone --depth 1 https://github.com/jcwleo/oh-no-harnessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/jcwleo/oh-no-harness/ultrawork)<a href="https://agentmods.dev/skills/jcwleo/oh-no-harness/ultrawork"><img src="https://agentmods.dev/badge/skills/jcwleo/oh-no-harness/ultrawork/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/jcwleo/oh-no-harness/ultrawork"><img src="https://agentmods.dev/badge/skills/jcwleo/oh-no-harness/ultrawork.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00034 | $0.05990 |
| Opus 5 | $0.00017 | $0.02995 |
| Sonnet 5 | $0.00007 | $0.01198 |
| Haiku 4.5 | $0.00003 | $0.00599 |
Grade A, and why
ultrawork scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 476 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ultrawork for Claude Code
This generated file is the Claude Code-facing runtime skill document. Claude Code slash commands should read this file directly; maintainers edit the source documents listed below instead.
Generated Runtime Composition
Source order:
../../docs/skill-core/ultrawork.md../../docs/platforms/claude-code-ultrawork.md
The sections below are already composed for this platform. Do not ask the runtime model to load another platform's runtime document or invocation syntax.
Source: docs/skill-core/ultrawork.md
Ultrawork
Ultrawork is a Markdown-first orchestration loop for moving from idea to
verified result with the retained skill chain. Each phase is chosen
explicitly from this Markdown workflow — there is no hidden next-step
selector. Do not use it for a small concrete fix whose contract surface,
baseline evidence, and verification command are already clear; use direct
implementation or ralph.
Invariants
U1. Ultrawork orchestrates the retained chain and never replaces it: the
planning gate uses `ralplan`, the execution handoff uses `ralph`.
When Ralph is unavailable, Ultrawork may inline the phase procedure but
remains an orchestrator: repository mutation still dispatches `executor` by
default under Ralph's need test; inline writes require a recorded LIGHT-tiny
or dispatch-unavailable fallback and carry the unchanged executor contract
per Ralph's `## Mode-Gated Agent Dispatch`, and a fired review or audit
trigger is exempt from the need test and never runs inline.
U2. Approved existing specs or plans may skip earlier phases only when the
skip reason and source artifact are recorded; a merely relevant plan
without approval evidence or with mismatched scope goes through the
planning gate.
U3. requirements_gate: planning must not start until the requirements
source is recorded. Interview is the only user-facing content approval
gate for new work; ultrawork never auto-approves the interview spec.
U4. Once ralplan's gates pass, record the plan approval source and
continue into ralph without a separate Plan Approval Brief prompt;
pause only on ralplan's named pause conditions (canonical list:
ralplan `### Ultrawork exception`).
U5. worktree_gate: no source file edit until a `Worktree decision` is
recorded. Ultrawork's worktree duties are the numbered list in
`### WORKTREE` below.
U6. The Direction Contract and AC IDs carry unchanged through plan,
session, packets, and report; each phase records only its delta; a
phase needing a direction change pauses for explicit user approval.
U7. Markdown at .oh-no/sessions/{sessionId}/ultrawork.md is authoritative.
No JSON state artifact in v1; any future JSON must be derived and
non-authoritative. No timer, daemon, or background heartbeat.
U8. Doctor/status runs at entry, resume, pre-execution, pre-merge, and
pre-final; BLOCKED stops before edits, merge, or final claim.
U9. Escalation routes are fixed: ambiguous requirements -> user or
`interview`; direction/scope conflict -> user or `ralplan`; failing
checks or unknown root cause -> `systematic-debugging` (which owns
`debugger` dispatch); public-contract/security/packaging risk ->
`code-reviewer` or `verifier`; missing worktree or verification
evidence -> blocked.
U10. Any scope change, missing authority artifact, failed worktree gate,
or failed verification transitions to a named non-success outcome,
never silent continuation.
U11. Final Validation does not repeat Ralph's completed internal gates;
review-then-verify order holds and an early verifier is stale,
discarded, and rerun.
U12. The run is invalid unless the ledger shows each named phase gate
satisfied or a recorded not-required reason; a silently omitted step
is a named ledger gap, not a pass.
U13. Ultrawork is the only context that may invoke interview/ralplan/ralph
without their per-step transition question. It skips the between-phase
"which next skill?" question and the separate plan-approval prompt —
and skips nothing else: spec approval when requirements are unclear,
planning gates, scope-change pauses, verification, and final evidence
all still run.
U14. Maker roles do not self-approve; inline checker fallback is still
checker output. When a named verifier trigger fires on a
subagent-capable host, an
inline check by the maker or accepting agent never satisfies the
independent verifier audit. Same authorship alone is not that trigger.
U16. Phase-agent authorization, eligibility, and fallback follow
`## Agent Roles`.
U17. A discovered plan admitted through U2 but lacking an execution
profile gets the profile set before execution (schema is
ralplan-owned); execution never starts profile-less.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 476 lines · 34 tokens per session scan A 6c8eb8b1c58a
ultrawork is a skill published in the GitHub repository jcwleo/oh-no-harness (11 stars, last pushed 1mo ago), licensed MIT. It adds 34 tokens to every session and 5,990 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
insight-error-page
Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…
next-cache-components-optimizer
Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…
next-partial-prefetching-adoption
Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…