agentleFS
mblaubergmblauberg/provenant/skills/implement/SKILL.md

implement

Skill1 starsChanged 4 days agoMITView on GitHub

Use to build an approved software change through the deliver loop: write, CI, repair, independent review, land. Not for unsettled scope (scope), root cause (diagnose), read-only review (code-review) or promotion (release).

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 10 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

  • AWSSkill

    to-questionnaire

    Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

    Also in 88 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

  • GoogleSkill

    to-spec

    Turn the current conversation into a spec and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.

    Also in 168 projects that use this

  • +3 more from google

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 171 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

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

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

    Use implement for a slice and deliver for the full delivery loop.

    line 8
  2. Very specific:

    Start from an approved spec: outcome, acceptance criteria with how each is verified, non-goals, risk tier and write authority (the tracker …

    line 12
  3. Distinctive:

    Read repository checks, commit rules and review floors in AGENTS.md and runbooks.

    line 20
  4. Very specific:

    Run provenant run --route junior --file slice.toml --explain, then omit --explain to submit.

    line 36
  5. Very specific:

    Check authority, write scopes and gates in the explanation, then submit without --explain.

    line 39
  6. Very specific:

    Follow provenant status, collect provenant output TARGET, and use retry for unfinished work.

    line 45
  7. Very specific:

    Update the owned docs in the same PR: behaviour, architecture, operations.

    line 74
  8. Distinctive:

    Keep compatibility only for published or persisted contracts.

    line 76

“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

Use implement for a slice and deliver for the full delivery loop.

Entry

Start from an approved spec: outcome, acceptance criteria with how each is verified, non-goals, risk tier and write authority (the tracker issue or the scope digest). Missing inputs go back to scope. Do not invent owner decisions; record an engineering call as Decided (<source>): <reason> and proceed.

Show provenant tracker lint N advice before starting an issue unless already supplied; unavailable checks do not block work.

Read repository checks, commit rules and review floors in AGENTS.md and runbooks. Run targeted checks locally; leave full suites to CI.

How it runs

Register a writer worktree first (provenant worktree create slice --base BASE). Save slice.toml, replacing the path, scope and brief:

use = "implement"
tier = "junior"
worktree = "/absolute/project/.worktrees/slice"
write_scope = ["src"]
brief = "Meet the approved criteria; run the named checks, commit and report."

Run provenant run --route junior --file slice.toml --explain, then omit --explain to submit.

Check authority, write scopes and gates in the explanation, then submit without --explain. deliver is the default composite; use implement for one slice, implement.fix for a known defect, and batch for parallel slices into one integration branch. Use plan when the shape is open. See templates.

Follow provenant status, collect provenant output TARGET, and use retry for unfinished work. edit changes unstarted nodes. Landing uses provenant land N --head SHA --wait after the integrated head meets its gates.

Writing the brief

The writer sees only the brief. Put in it the issue path, the acceptance criteria, the files and owners already found, the checks to run, and the instruction to stop and report instead of widening scope. One writer per worktree. Name the method skill: test-design for observable behaviour, refactor for restructuring, diagnose for an unknown cause.

A separate fresh writer attempt repairs qualified review findings; the raising lenses re-review the delta. max_repair_cycles defaults to three and accepts one to three, independently of the CI repair limit. A recurring finding or an evidenced design defect escalates early to read-only principal scoping. Proceed within approved scope: the principal records repair_design booleans within_scope and one_way_door in its findings block. The host notifies the owner and bypasses its decide wait only for explicit approved scope with no one-way door. Changed scope, one-way doors and missing assessments wait on that host decision before any continuation writer. Approval retains the original literal write authority. A failed replanned panel feeds its qualified findings into the remaining repair cycles. If no separate route is available, continue in a fresh attempt on the available route with a recorded separation warning.

Writer rules

  • The implementer may not edit a failing acceptance test or weaken an assertion to pass. A changed test is reported with the reason.
  • Update the owned docs in the same PR: behaviour, architecture, operations. A settled decision moves to its durable home (see engineering-docs).
  • Keep compatibility only for published or persisted contracts. Otherwise change the callers and delete the old path.
  • Report checks that did not run as unverified, never as passing.
  • A repair that leaves failures unchanged or larger twice in a row stops the run. Return the evidence to scope when the spec turns out wrong.

Done

Exact-head CI and the review the tier requires meet acceptance. Finish with the outcome report owned by deliver. Promotion needs separate authority (release).

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.