agentleFS
tamesystemstamesystems/intent-engineering-skills/skills/pave-path/SKILL.md

pave-path

Skill0 starsChanged 39 days agoMITView on GitHub

Make a validated architectural path the easiest supported way to create or extend code by adding a generator, template, constructor, helper, public entry point, test harness, or other repository-owned affordance. Use when contributors repeatedly need to perform the same architectural operation or the correct path is harder than a known-invalid one.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Choose the smallest affordance: Prefer, depending on the problem: an obvious public API/constructor.

    line 15
  2. Distinctive:

    Prove the path: Use the affordance to create or exercise one representative instance, run the authoritative checks, and inspect the result.

    line 18
  3. Very specific:

    Pair with enforcement when justified: If an easy forbidden path still competes with the paved path, route the validated invariant to …

    line 19
  4. Distinctive:

    Explain discoverability: Ensure an unfamiliar contributor can find the paved path from local repository context, diagnostics, scripts, or …

    line 20

≈ “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

Pave Path

Make doing the right thing require less invention than doing the wrong thing. Constraints without an attractive supported path create thrashing; a pit of success needs both guidance and gravity.

Work

  1. Name the recurring operation. Examples: create a module, add a command, define a capability, add a migration, construct a domain value, register a route, or test a user-visible behavior.
  2. Confirm the architecture is validated. The path must represent a real, repeated or consequential convention. If ownership/boundary design is unsettled, route to design-change or design-module first; do not fossilize a guess in a generator.
  3. Observe the best existing instance. Recover the minimum structure, imports, verification, naming, and registration steps that make a native implementation correct.
  4. Choose the smallest affordance. Prefer, depending on the problem: an obvious public API/constructor; a copyable local example; a typed helper; a generator/codemod; a package template; a verification harness. Do not build a framework when a function or tiny script removes the decision.
  5. Encode only stable decisions. Generate/abstract the parts contributors should not have to reinvent. Leave product/domain choices explicit rather than filling them with fake defaults.
  6. Make output repository-native. Generated code should use existing naming, package manager, formatting, tests, types, public surfaces, and check commands. It must not introduce a parallel architecture.
  7. Prove the path. Use the affordance to create or exercise one representative instance, run the authoritative checks, and inspect the result. For generators/codemods, prefer idempotence or a safe refusal over destructive reruns.
  8. Pair with enforcement when justified. If an easy forbidden path still competes with the paved path, route the validated invariant to harden-boundary, create-check, or ratchet.
  9. Explain discoverability. Ensure an unfamiliar contributor can find the paved path from local repository context, diagnostics, scripts, or concise navigation documentation.

Return

  • Operation paved
  • Affordance created/improved
  • Decisions encoded
  • Decisions deliberately left open
  • Proof it produces native valid work
  • Remaining competing footguns

Failure signals

Watch for generating speculative architecture, giant scaffolds full of placeholders, a new tool where an existing API suffices, hiding meaningful product decisions behind defaults, generators that drift from the repository's checks, and prohibiting a path without making the replacement easy to discover.

Done when

A representative contributor can perform the recurring operation through a discoverable repository-owned path that produces valid native structure with less architectural invention than doing it manually, and remaining cheap competing footguns are identified for ratcheting.

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.