agentsclimarketplace

Workflow

Skill Stepkar2004/init-configurator/.claude/skills/workflow

Agentic project base: Claude Code skills bootstrap and evolve your project, a deterministic Python CLI verifies it. project.yaml manifest, doctor checks, path linting. A project-template alternative for coding agents.

Install
npx -y skills add Stepkar2004/init-configurator --skill workflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

The SWE loop binding every implementation session in this repo - fire whenever code is about to be written, changed, or resumed, at every chunk start and end, and whenever unsure what the next step is. Covers orient -> plan -> implement -> verify -> document -> commit (NEVER push) -> reflect. Nested lazy parts; load references/scale.md when work outgrows one context window (massive task/refactor, flood of tasks, mass research); load references/rot-check.md before a release, after a long repo gap, or on "rot check"/"is anything stale". Not for scaffolding new stacks (bootstrap) or skill hygiene (skill-manager).

SKILL.md

3.6 KB, 733 tokens by cl100k_base, as published. Nobody here has run it

workflow — the loop every implementation session runs

One skill fires for all SWE work; the less-common parts load lazily from references/ when their trigger row matches. Do not rely on separate sibling skills auto-triggering for scale or rot-check — this file is their front door.

The loop

  1. Orient. Read docs/state/now.md, then the active plan (docs/plans/ when the repo keeps phase plans, else the current chunk in docs/state/roadmap.md). Know the chunk's goal and acceptance criteria before touching code. If no chunk is defined, defining one IS the first task.
  2. Plan the chunk. Smallest end-to-end slice; name the files to touch and the tests that will prove it. If scope grows mid-chunk beyond the plan, stop — re-plan or split; never "just keep going".
  3. Implement. Follow the constitution (CLAUDE.md) and the architecture rules in the active plan — they are constraints, not suggestions.
  4. Verify — the forgettable steps, in order:
    • New behavior ⇒ new test; changed behavior ⇒ changed test. Same chunk, no IOUs.
    • Run the repo's gates: project.yaml declares them as tasks; the constitution (CLAUDE.md) and the pre-commit / CI configs are the full set.
    • User-visible behavior ⇒ actually run it — a green unit test is not a seen behavior.
  5. Document in the same chunk. Update whatever the change made stale: README, the plan doc, an ADR for any new decision (docs/decisions/), docstrings. The test: "would a fresh session mis-learn anything if it read the docs right now?"
  6. Commit at the chunk boundary. Gates green → docs/state/now.md + log.md updated → commit. NEVER push. The user pushes, or explicitly says push — this repo is public; an unpushed mistake is free, a pushed one is not.
  7. Reflect — do not skip, do not defer. Ask it literally: did a lesson land this chunk (a root cause found, a tool swapped, a convention decided, a default overridden, a mistake repeated)? If yes, run the evolve procedure NOW (skill-manager/references/evolve.md) before the session ends — a lesson you plan to record "later" is a lesson lost, and re-teaching it next session is the tax. Task matched no skill? → one line in docs/state/miss-log.md, keep working.

When to stop and ask

The user is available and asking is free: scope changes, tradeoffs with product impact, anything irreversible or public-facing, or two defensible designs with different long-term costs. Blocked beats wrong.

Lazy parts (load only on trigger)

TriggerLoad
Work outgrows one context window: massive task or refactor, flood of small tasks, mass research, quality degrading with context sizereferences/scale.md
Before a release; after a long gap in the repo; "rot check" / "is anything stale"; a gate that has not been seen failing latelyreferences/rot-check.md

What ships with it: 2 files

5.3 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 327,132. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.