agentleFS
tmussertmusser/ai-engineering-skills/skills/ceremony-budget/SKILL.md

ceremony-budget

Skill6 starsChanged 3 months agoMITView on GitHub

Choose the smallest workflow that still buys back enough attention and safety for the task.

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 13 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use when the safe route is not already obvious.

    line 16
  2. Distinctive:

    Check whether the task is bounded, reversible, and already explicit.

    line 29
  3. Distinctive:

    Start the next action immediately once the boundary and proof target are clear.

    line 34
  4. Distinctive:

    create process work whose only purpose is documenting the process choice

    line 46
  5. Distinctive:

    When an existing route is sufficient, prefer:

    line 48
  6. Distinctive:

    If one guard is missing, provide only that delta:

    line 54
  7. Distinctive:

    Do not escalate only because the repository is large, many skills are installed, or a fuller route is available.

    line 88
  8. Very specific:

    When two routes protect the same named risks, an available Ceremony preference may choose between them: lean favors the lighter route, …

    line 96
  9. Distinctive:

    Use at most this six-line block, and omit it when the existing route is enough:

    line 104
  10. Very specific:

    Do not create CEREMONY_BUDGET.md or a budget ledger by default.

    line 115
  11. Very specific:

    Use handoff only when another session actually needs to resume, work remains unresolved, or durable continuation state buys real safety.

    line 125
  12. Distinctive:

    Stop when all are true:

    line 139
  13. Distinctive:

    no named acceptance criterion or risk remains unresolved

    line 143
  14. Distinctive:

    After that, do not rerun an already passing proof validator without a concrete reason, repeat Git or status commands for reassurance, …

    line 145
  15. Distinctive:

    After a targeted check fails:

    line 152
  16. Distinctive:

    Do not respond to one failed check by generating additional planning artifacts.

    line 161

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Ceremony Budget

Purpose

Spend ceremony only when it buys back attention or safety. Route a task to a workflow that replaces the default larger route; do not append a second process contract.

When to use

Use when the safe route is not already obvious. If the task or wrapper already supplies an execution-sized route, use it without re-deriving the route.

Inputs

  • User request and any task-wrapper instructions
  • Scope boundary and expected blast radius
  • Verification target and failure risk
  • Ambiguity, irreversibility, and resume risk
  • Optional operator profile defaults

Workflow

  1. Check whether the task is bounded, reversible, and already explicit.
  2. Identify the concrete implementation, reproduction, or inspection action and its proof target.
  3. Choose the lowest level that protects the actual failure mode.
  4. Select one replacement route and name only the missing guard, if any.
  5. Start the next action immediately once the boundary and proof target are clear.

Replacement rule

Ceremony budget selects a route that replaces the default larger workflow. Do not regenerate the full route after a wrapper has already made it explicit. It must not:

  • append a second workflow contract to an already explicit route
  • restate task-wrapper instructions
  • invoke several skills merely because they are available
  • require a budget ledger merely to prove ceremony was budgeted
  • create process work whose only purpose is documenting the process choice

When an existing route is sufficient, prefer:

Route already explicit; no additional ceremony block needed.

If one guard is missing, provide only that delta:

Route already explicit.
Missing guard: stop after verification and required proof pass.
Next action: implement the named seam.

Startup discipline

For Level 0 and Level 1, do not create a durable planning artifact by default or run commands solely to document route selection. Identify the concrete action, then begin it once the boundary and proof target are clear.

For Level 2, allow one compact acceptance block only when real ambiguity exists. Implementation must follow immediately; do not create secondary artifacts before the first implementation attempt.

Level 3 may retain fuller safeguards when the risk genuinely warrants them.

Fast routing test

Evaluate these from lightest to heaviest and choose the first level that leaves no named risk uncovered:

  • Level 0: the edit is tiny, reversible, behaviorally obvious, and one sanity check can prove it.
  • Level 1: one bounded behavior seam changes, acceptance is already clear, and one targeted verification path can prove it.
  • Level 2: acceptance criteria, compatibility seams, data meaning, or test meaning must be written down before implementation can be trusted.
  • Level 3: the work is multi-slice, autonomous, hard to reverse, user-facing, decision-impacting, data-sensitive, or unsafe to resume without durable state.

Do not escalate only because the repository is large, many skills are installed, or a fuller route is available. Escalate because a lighter route leaves a concrete risk unprotected.

Operator-profile tie-breaker

Operator profile is a tie-breaker, not a risk override.

When two routes protect the same named risks, an available Ceremony preference may choose between them: lean favors the lighter route, balanced keeps the normal choice, and guarded may favor slightly more durable evidence when the extra cost is small. Never use the profile to downgrade required proof, skip a named risk, or escalate ceremony without a concrete benefit.

Outputs

Use at most this six-line block, and omit it when the existing route is enough:

CEREMONY ROUTE
Level: 0 | 1 | 2 | 3
Next action: ...
Use: ...
Skip: ...
Proof and stop: ...

Do not create CEREMONY_BUDGET.md or a budget ledger by default.

Level guidance

  • Level 0: direct patch -> one sanity check -> stop
  • Level 1: inline boundary -> build-one -> targeted verify -> stop
  • Level 2: compact mini-spec -> build-one -> targeted test -> verify-contract -> stop
  • Level 3: fuller guarded route for genuinely high-risk, multi-slice, or decision-impacting work

Use handoff only when another session actually needs to resume, work remains unresolved, or durable continuation state buys real safety. It is not automatic.

Compatibility reserve

For each named preserved behavior, non-goal, or adjacent compatibility seam, reserve the cheapest relevant negative or regression check when one is available.

Light routing may omit planning artifacts. It must not omit a cheap check that protects an explicitly preserved behavior. Reserving that check is proof work; it does not by itself require a larger route.

Stop conditions

Stop when all are true:

  1. required verification has passed
  2. required proof is valid
  3. no named acceptance criterion or risk remains unresolved

After that, do not rerun an already passing proof validator without a concrete reason, repeat Git or status commands for reassurance, rewrite valid artifacts for style, continue because turns remain, add a ceremony retrospective, or expand verification beyond the named risk without evidence that it is needed.

Failure discipline

After a targeted check fails:

  1. inspect the first actionable traceback or mismatch
  2. inspect the actual import, call, or data path implicated by the failure
  3. make the smallest related correction
  4. rerun the narrow check
  5. escalate ceremony only if the failure reveals genuine ambiguity, broader scope, or higher risk

Do not respond to one failed check by generating additional planning artifacts.

Anti-patterns

  • Treating more artifacts as automatically better process.
  • Making the six-line block mandatory when no route delta is needed.
  • Deriving a full route after a wrapper has already made it explicit.
  • Requiring scope-freeze, handoff, a ledger, or a durable artifact for every Level 1 task.
  • Escalating because a large bundle is installed rather than because risk requires it.
  • Omitting a cheap regression check for an explicitly preserved behavior.
  • Stopping before required proof is valid.
  • Using Level 0 or 1 when the task is decision-impacting or genuinely unsafe to resume without durable state.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.