live-money-discipline
Skill0 starsChanged 8 days agoMITView on GitHub
Use when building or changing a system whose actions are irreversible — moving money, trading, production deploys, data deletion, external sends. Gate every live action behind dry-run, explicit go-live signoff, a kill switch, and blast-radius limits, and default to the reversible path.
In this repository108 other files it gives its agents.
- adaptive-estimationSkill
- adaptive-loop-budgetingSkill
- adopt-and-deleteSkill
- adversarial-reviewSkill
- agent-drives-uiSkill
- agentic-coding-loopSkill
- agent-otel-trajectorySkill
- agents-as-iac-modulesSkill
- ascii-ui-mockupsSkill
- authoring-html-emailSkill
- authoring-skills-vs-mcpSkill
- batched-ssh-enumerationSkill
- c4-architectureSkill
- challenge-the-premiseSkill
- codegraph-bootstrapSkill
- codegraph-worktree-startupSkill
- colocated-pod-tolerationsSkill
- command-creatorSkill
- complexity-budgetsSkill
- connector-readinessSkill
- context-anchored-patchingSkill
- continuous-issue-resolutionSkill
- credential-proxySkill
- criticality-tiered-reportingSkill
- data-connector-buildingSkill
- declarative-firstSkill
- disk-pressure-forensicsSkill
- early-return-over-elseSkill
- email-triageSkill
- enums-codify-behaviorSkill
- ephemeral-workspace-lifecycleSkill
- event-sourcingSkill
- evidence-claim-boundariesSkill
- exhaustive-matchSkill
- feature-coverage-not-just-line-coverageSkill
- full-disk-forensicsSkill
- gate-before-pushSkill
- github-pr-shepherdingSkill
- graceful-shutdown-stateful-agentsSkill
- grounded-generationSkill
- iac-incident-responseSkill
- iac-not-ad-hocSkill
- instrumented-service-scaffoldSkill
- interface-over-conditionalsSkill
- knowledge-routingSkill
- lean-on-oss-standardsSkill
- live-state-verificationSkill
- loop-governance-and-learningSkill
- low-level-executor-task-specSkill
- map-dispatch-over-conditionalsSkill
- merge-conflict-disciplineSkill
- mermaid-diagramsSkill
- oke-arc-runners-opentofuSkill
- orchestrate-agySkill
- orchestrate-claudeSkill
- orchestrate-codexSkill
- orchestrate-geminiSkill
- otel-emit-at-willSkill
- parameter-designSkill
- parse-dont-validateSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 35
Before claiming a live change safe, verify the actual artifact in the real environment (verify-real-artifact) and have an independent …
- Distinctive: line 45
Treat the runtime risk knobs (limits, kill switch, mode) as config you can change live without a deploy, so reacting to trouble is seconds, …
- Distinctive: line 50
If the only thing between a typo and an irreversible effect is remembering to be careful, the system is wrong — add a gate.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Live-Money Discipline
Some actions can't be taken back: a trade fills, money leaves, a row is deleted, an email sends, a bad image rolls to the fleet. For these, "we'll fix it after" doesn't exist — so the engineering bar is different. The discipline: make the irreversible action the hardest thing to do by accident, and the reversible path the default. Money is the sharpest example; the rules apply to any irreversible effect.
The gates (each independently)
- Dry-run / paper mode first. The same code path must be able to run without committing the effect — log "would have done X" — and that mode is the default until explicitly armed. Most live-action bugs are caught here for free.
- Explicit go-live signoff for irreversible scope. Crossing into real money / production scale is a deliberate, recorded human decision, not a config default or a flag that ships on by accident.
- A kill switch that actually stops it. One runtime control (no redeploy) halts new live actions immediately, and you've tested that it does. A safety control you haven't exercised is decoration.
- Blast-radius limits / capital tiering. Cap the per-action and cumulative exposure (position size, batch size, rate, spend). Separate "can lose it all" risk capital from protected capital; never let one bug reach everything.
- Idempotency + reconciliation. Make retries safe (idempotency keys) and reconcile intended-vs-actual continuously, because at-least-once delivery means the action will sometimes fire twice.
Verify the real thing, adversarially
Before claiming a live change safe, verify the actual artifact in the real environment (verify-real-artifact) and have an independent reviewer try to refute it (adversarial-review) — the cost of a wrong "it's fine" here is realized, not hypothetical. Default to REVISE under uncertainty.
Defaults that bias to safety
- Reversible path is the default; irreversible requires arming.
- Fail closed — on error or ambiguity, don't take the live action.
- Make the dangerous command harder to issue than the safe one, not symmetric.
- Treat the runtime risk knobs (limits, kill switch, mode) as config you can change live without a deploy, so reacting to trouble is seconds, not a release.
Smell
If the only thing between a typo and an irreversible effect is remembering to be careful, the system is wrong — add a gate. Care is not a control.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

