Contract for hands-off mode — the ultrapack workflow variant that minimizes user prompts after Design, takes the safest reversible path, and logs every auto-choice for one end-of-task review. Referenced by /up:make and the child skills (udesign, uplan, uexecute, uverify, ureview). Read when the task file's Mode: header is hands-off.
EnglishUse to babysit a long-running process (training run, batch job, migration, deploy) on a remote pod or locally while the user is away. Defines a launch contract, gates on immediate crash, polls for stability, triages anomalies as recoverable (fix + resume) or not (tear down + notify). Auto-triggers when the user asks you to launch a job and watch it / keep it alive / shut it down if it dies.
EnglishUse on any bug, test failure, or unexpected behavior before proposing a fix. Enforces four-phase root-cause investigation — reproduce, pattern-match, hypothesize, fix — with a hard rule against symptom patches.
EnglishUse before any creative work — features, components, behavior changes. Turns an idea into a validated spec with explicit tradeoffs and unknowns, and splits scope into multiple tasks if it's too large. Output is the Design section of the task file.
EnglishUse when writing or editing documentation — project docs, CLAUDE.md, READMEs, SKILL.md, docstrings, inline comments. Auto-triggers on .md files and docstring edits. Enforces lead-with-why, kill stale content, lists over tables, no aspirational sentences.
EnglishUse to implement an approved plan. When the plan declares Interface graph, derives waves by topo-sort over [blocks] IF edges only — non-blocking IF edges (the default) put producer and consumer in the same wave, since the IF declaration is sufficient context. Edits inline when only one implementer would fire (single phase or serial fallback). Runs plan-diff check and consistency sweep after each commit. Forbids silent fallbacks and mutation of external spec/design docs. Dispatches the planner skill when deviations invalidate the plan.
EnglishUse after design to turn a validated spec into a lean implementation plan. Outputs the Plan section of the task file — files, line numbers, class/method names, invariants, test strategy, order. Ends with a scope-creep / simpler-way check.
EnglishUse after verify passes for the future maintainer's audit — sit in the chair of the person who'll touch this code in 3-6 months and ask "what will bite us later?" at the decision level. Surfaces wrong abstractions, load-bearing-but-unobvious shapes, next-change traps, drift from surrounding code; raises a Scope flag if the whole change looks like the wrong call. Dispatches up:reviewer (critical, high-confidence filter), processes findings fairly, fills the task file's Conclusion.
EnglishUse after execute to attack the change and demonstrate how it's broken. Default stance — the change is broken; prove and demonstrate it. Builds an attack checklist (happy-path / negative / invariant / interface hypotheses), runs each freshly, smokes the end-to-end, writes a short summary to the task file's Verify section, loops back to execute on any demonstrated break.
EnglishIncludes Job Guardian and other workflows.
Supported by the Atlas install prompt. Skill-specific requirements vary.
Pack stats
What’s included
Handsoff
Contract for hands-off mode — the ultrapack workflow variant that minimizes user prompts after Design, takes the safest reversible path, and logs every auto-choice for one end-of-task review. Referenced by /up:make and the child skills (udesign, uplan, uexecute, uverify, ureview). Read when the task file's Mode: header is hands-off.
Job Guardian
Use to babysit a long-running process (training run, batch job, migration, deploy) on a remote pod or locally while the user is away. Defines a launch contract, gates on immediate crash, polls for stability, triages anomalies as recoverable (fix + resume) or not (tear down + notify). Auto-triggers when the user asks you to launch a job and watch it / keep it alive / shut it down if it dies.
Udebug
Use on any bug, test failure, or unexpected behavior before proposing a fix. Enforces four-phase root-cause investigation — reproduce, pattern-match, hypothesize, fix — with a hard rule against symptom patches.
Udesign
Use before any creative work — features, components, behavior changes. Turns an idea into a validated spec with explicit tradeoffs and unknowns, and splits scope into multiple tasks if it's too large. Output is the Design section of the task file.
Udocument
Use when writing or editing documentation — project docs, CLAUDE.md, READMEs, SKILL.md, docstrings, inline comments. Auto-triggers on .md files and docstring edits. Enforces lead-with-why, kill stale content, lists over tables, no aspirational sentences.
Uexecute
Use to implement an approved plan. When the plan declares Interface graph, derives waves by topo-sort over [blocks] IF edges only — non-blocking IF edges (the default) put producer and consumer in the same wave, since the IF declaration is sufficient context. Edits inline when only one implementer would fire (single phase or serial fallback). Runs plan-diff check and consistency sweep after each commit. Forbids silent fallbacks and mutation of external spec/design docs. Dispatches the planner skill when deviations invalidate the plan.
Uplan
Use after design to turn a validated spec into a lean implementation plan. Outputs the Plan section of the task file — files, line numbers, class/method names, invariants, test strategy, order. Ends with a scope-creep / simpler-way check.
Ureview
Use after verify passes for the future maintainer's audit — sit in the chair of the person who'll touch this code in 3-6 months and ask "what will bite us later?" at the decision level. Surfaces wrong abstractions, load-bearing-but-unobvious shapes, next-change traps, drift from surrounding code; raises a Scope flag if the whole change looks like the wrong call. Dispatches up:reviewer (critical, high-confidence filter), processes findings fairly, fills the task file's Conclusion.
Uverify
Use after execute to attack the change and demonstrate how it's broken. Default stance — the change is broken; prove and demonstrate it. Builds an attack checklist (happy-path / negative / invariant / interface hypotheses), runs each freshly, smokes the end-to-end, writes a short summary to the task file's Verify section, loops back to execute on any demonstrated break.
Quick start
- Choose your skills. Select individual skills above, or choose the entire pack.
- Copy the install prompt. Use Copy install prompt above the list.
- Send it to your agent. Paste into your coding agent’s chat. Your agent reviews the source and guides the setup.
Try it with your agent
After installation, start with a request like:
Use the Handsoff skill to help with my task.
Ask me for the context you need before getting started.Requirements & license
- A coding agent that can access your project files and install skills.
- Your agent checks each skill’s instructions, dependencies, and permissions during setup.
- Reported license for the skills shown: WTFPL. The author’s terms apply.