agentleFS
siarhei-belavussiarhei-belavus/agent-smitt/skills/engineering/implement/SKILL.md

implement

Skill4 starsChanged 15 days agoMITView on GitHub

Implement repository deliveries from authoritative specs or tickets. Use for standalone implementation and for one repository delivery narrowed by deliver-ticket.

191 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-ts-deep-modules, ask-matt and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 43 projects that use this

  • GoogleSkill

    ask-matt

    Ask which skill or flow fits your situation. A router over the skills in this repo.

    Also in 91 projects that use this

  • RoentekSkill

    loop-me

    Grill me about specs for the workflows I want to build, within this workspace.

    Also in 50 projects that use this

  • ryotaroh180105Skill

    claude-handoff

    Hand the current conversation off to a fresh background agent that picks up the work immediately.

    Also in 42 projects that use this

  • LesaniSkill

    writing-beats

    Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.

    Also in 47 projects that use this

  • AWSSkill

    implement-spec

    Implement the result of /to-spec and /to-tickets in code.

    Also in 67 projects that use this

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 8 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 171 edited versions we read keep these.

  • 8

    Distinctive

    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:

    Read delivery records before resolving the assignment.

    line 8
  2. Distinctive:

    When the user authorizes one or more writable Repository Deliveries directly, read Standalone assignments.

    line 14
  3. Very specific:

    When deliver-ticket supplies one complete narrowing assignment for exactly one Repository Scope entry, read Coordinator narrowing.

    line 15
  4. Distinctive:

    Every writable Repository Delivery must supply or resolve:

    line 17
  5. Distinctive:

    Use the authorized isolated writable checkout.

    line 39
  6. Distinctive:

    Run focused tests and typechecking during the work.

    line 49
  7. Distinctive:

    Complete when: every authorized outcome is implemented in committed green slices, classified unchanged, or represented by an exact blocker …

    line 51
  8. Distinctive:

    Run the full repository-owned validation suite after implementation.

    line 59
  9. Distinctive:

    Complete when: every authorized Repository Delivery satisfies its assignment's completion requirements and the requested result has been …

    line 71

“Common” is measured: kept by at least 6 in 10 of the 171 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

Implement the work described by the user in the authoritative spec or tickets.

Read delivery records before resolving the assignment.

Resolve the assignment

Classify the assignment before Context Assembly or checkout changes.

  • When the user authorizes one or more writable Repository Deliveries directly, read Standalone assignments.
  • When deliver-ticket supplies one complete narrowing assignment for exactly one Repository Scope entry, read Coordinator narrowing.

Every writable Repository Delivery must supply or resolve:

  • Repository ID and Repository Reference;
  • writable checkout or Execution Worktree and its branch;
  • exact launch commit or Fixed Review Base;
  • repository-specific required outcome;
  • authoritative sources and repository instructions;
  • complete Settled Seam records and their test approaches;
  • repository-owned Validation Obligations and the repository sources that define validation commands.

A Fixed Review Base is the exact starting commit against which a Repository Delivery is implemented and reviewed.

Read-only context and validation repositories remain outside writable scope. Missing authority, a worktree that disagrees with the assignment, or a required change to approved scope, acceptance behavior, or a Settled Seam is a Material contradiction. Preserve safe work and return the conflicting sources. Authority comes from the assignment, not inference.

Complete when: the assignment mode is explicit; the selected branch's authority gate passes; the writable scope, authority, and every required delivery field are complete; or a Material contradiction or explicit authority stop has been returned without repository mutation.

Prepare each delivery

Validate that the writable path belongs to the Repository Reference and that its starting revision agrees with the supplied launch commit or Fixed Review Base. Read that repository's instructions and resolve its validation commands from the supplied sources. Before each implementation decision, perform configured Domain Orientation for the affected behavior and read the applicable canonical context, ADRs, and nearest relevant prior art. Use the prepared ticket's self-contained contract; follow indexed sources when that decision needs them. Do not repeat publication's planning audit or use one repository's instructions as authority for another.

Record the Repository ID, writable path, Fixed Review Base, authoritative sources, complete Settled Seams, test approaches, Validation Obligations, and validation commands before editing.

Use the authorized isolated writable checkout. Classify existing changes against the assignment. Preserve unrelated content and return a blocker when it cannot be isolated without changing user work.

Complete when: the writable path, starting revision, repository instructions, authorized outcomes, seams, and validation approach permit the first implementation slice, and unrelated content is either isolated or reported as a blocker. Required domain reading precedes the slice that depends on it.

Implement through TDD

Treat each task as an increment to the application's accumulated invariants. Accepted task decisions authorize adding, changing, or removing specific invariants; existing ones otherwise remain in force. Before removing or weakening behavior whose purpose is unclear, investigate the invariant it protects through surrounding code, tests, and relevant commit history. Long-standing code warrants extra scrutiny, though age does not prove correctness. If its purpose remains unclear or conflicts with the task, preserve it and escalate with evidence and the concrete tradeoff. Proceed when an accepted decision authorizes the change or evidence establishes an implementation error within scope. Satisfying the new requirement alone does not justify erasing an existing invariant.

Work in vertical slices. Use the tdd skill with the authoritative sources and complete Settled Seam records, including each seam's owning module, caller-visible interface, location, status, observable behavior, test approach, and nearest prior art.

Run focused tests and typechecking during the work. After each coherent green iteration, commit focused changes to that delivery's branch. Each commit contains changes from one repository only.

Complete when: every authorized outcome is implemented in committed green slices, classified unchanged, or represented by an exact blocker or Material contradiction.

Validate the exact result

Ad hoc validation is allowed. Keep one-off validation code outside the committed delivery. If a check belongs in the MR or PR, implement it as a regular test following the repository's existing test conventions. Do not commit ad hoc runners or their supporting configuration and documentation.

Establish a Clean delivery state before final validation: every assignment-owned change is committed and the worktree is clean. Isolate unrelated or uncommitted content safely, or return a blocker with the exact status. Validation from a dirty worktree is not attributable to HEAD.

Run the full repository-owned validation suite after implementation. Record every command, outcome, and relevant limitation against the exact commit it validated. Resolve the Exact local HEAD and clean-worktree status after validation. A changed head or worktree makes prior validation evidence stale and requires a refresh.

Return one Repository Delivery result for each writable entry with:

  • Repository ID, Repository Reference, writable path, and branch;
  • Fixed Review Base and Exact local HEAD;
  • changed, unchanged, or blocked outcome;
  • Clean delivery state;
  • focused commits;
  • repository validation commands and outcomes;
  • remaining blockers or Material contradiction details.

Complete when: every authorized Repository Delivery satisfies its assignment's completion requirements and the requested result has been returned.

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.