agentleFS
tamesystemstamesystems/intent-engineering-skills/skills/ratchet/SKILL.md

ratchet

Skill0 starsChanged 40 days agoMITView on GitHub

Convert a recurring or consequential engineering failure, correction, or fragile convention into the strongest cheap durable repository mechanism. Use when a lesson should become harder or impossible to violate rather than another instruction.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/ratchets.md and references/mechanism-ladder.md before choosing a mechanism.

    line 10
  2. Distinctive:

    Choose the strongest cheap mechanism: Use the mechanism ladder. Prefer prevention over detection and detection over instruction.

    line 17
  3. Very specific:

    Route implementation: Use harden-boundary for dependency topology, create-check for deterministic rejection, wire-check for authoritative …

    line 19

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “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

Ratchet

Turn engineering experience into accumulating executable constraints. This is a meta-engineering skill, not a ninth intent operator: it changes the environment future realize and verify work executes inside.

Read references/ratchets.md and references/mechanism-ladder.md before choosing a mechanism.

Work

  1. Observe the evidence. Name the concrete failure/correction and whether it is repeated, costly, consequential, or likely enough to justify permanent machinery. Do not invent a standard from taste alone.
  2. Generalize the failure class. State the invariant independently of the observed file or incident. Distinguish a mechanical property from a judgment call.
  3. Interview the repository. Find its existing compiler, linter, architecture checks, tests, generators, umbrella check command, CI, and local documentation. Prefer extending existing machinery over adding a new tool.
  4. Choose the strongest cheap mechanism. Use the mechanism ladder. Prefer prevention over detection and detection over instruction. Before creating a custom mechanism, check whether the existing toolchain or a well-established compatible executable already implements the invariant.
  5. Keep the good path cheap. Identify the intended alternative. If the new prohibition would leave contributors guessing, improve the canonical API/type/helper/generator/diagnostic or add a narrow navigation pointer.
  6. Route implementation. Use harden-boundary for dependency topology, create-check for deterministic rejection, wire-check for authoritative execution, and harden-check for meaningful bypasses. A single small ratchet may be implemented directly when delegation would add ceremony.
  7. Prove and report. A mechanical ratchet is not complete until a representative recurrence is rejected and the clean repository passes again. Important checks must be wired into the authoritative path.

Return

  • Outcome: RATCHETED / DOCUMENTED_JUDGMENT / NOT_WORTH_RATCHETING / BLOCKED
  • Observed lesson
  • Generalized invariant
  • Mechanism chosen — and why stronger/cheaper alternatives were unsuitable
  • Installed change
  • Positive path
  • Negative proof
  • Authoritative path — where the check runs
  • Remaining exposure

Failure signals

Correct course if you add prose without testing stronger mechanisms, add a new dependency when existing tooling can express the rule, create a prohibition without a usable alternative, declare success because clean code passes without proving the rule bites, or overfit a permanent constraint to one harmless anomaly.

Done when

The lesson has a justified durable disposition and, for a mechanical ratchet, repository-owned enforcement has been proven against a representative violation and appropriately wired into normal verification.

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.