agentleFS
RafaelFerSilvaRafaelFerSilva/refinement-agent-skills/skills/configure-refinement/SKILL.md

configure-refinement

Skill2 starsChanged 14 days agoMITView on GitHub

Configures a workspace for refinement workflows: operating mode (with or without code), local markdown tracker, publishing conventions, and agent instructions block in AGENTS.md or CLAUDE.md. Run once per workspace.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Write the workspace-level configuration that /bootstrap-context, /refine-story, /qa-plan, /requirements-discovery, and /meeting-minutes …

    line 9
  2. Very specific:

    Work items: Do User Stories originate in an external tracker?

    line 23
  3. Very specific:

    New product: if CONTEXT.md does not exist yet, no action is needed now: the first examination that resolves a domain term will create it.

    line 26

≈ “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

Configure workspace

Write the workspace-level configuration that /bootstrap-context, /refine-story, /qa-plan, /requirements-discovery, and /meeting-minutes assume. This is a conversation-guided skill, not an automated script: explore, present findings, confirm, and only then write files.

Steps

  1. Explore. Inspect existing files before making assumptions: AGENTS.md and CLAUDE.md in the root, CONTEXT.md, docs/agents/refinement.md and docs/agents/issue-tracker.md (previous outputs of this skill, including any codebase path(s) already listed under ## Fact sources), docs/product-overview.md, docs/reference/ and other documents in docs/, .scratch/, and signals of source code in the current workspace root (package.json, pyproject.toml, src/, go.mod, or anything a team compiles or executes). If docs/agents/issue-tracker.md already exists from another convention, preserve it: verify only that it specifies where to publish. If docs/agents/refinement.md already exists, accept both the canonical mode tokens (with-code, without-code) and the legacy tokens (com-codigo, sem-codigo) on read. If CLAUDE.md is merely a bridge with @AGENTS.md, the primary instruction file is AGENTS.md. Completed when: you know whether the workspace is already configured, the likely mode (with-code or without-code), any previously configured codebase path(s), the likely language (en-US or pt-BR), whether the product appears new or already in production, whether CONTEXT.md, docs/product-overview.md, or reference documents in docs/ exist, whether the tracker exists, and which instruction file to edit.
  2. Inquire, one section at a time, leading with the recommended answer.

    Mode and codebase paths. Recommend based on findings: if code is present in the workspace root, recommend with-code (defaulting to . and asking whether additional local repositories also belong to the product, such as sibling backend, frontend, or mobile repos); if code is absent from the workspace root, recommend without-code when no local codebase checkout is available, while noting that with-code also works from a dedicated refinement workspace by pointing to one or more local repository paths (for example ../product-api and ../product-web). Explain in one line: with code, the agent inspects source code read-only before asking; without code, the PO is the source of facts and architecture questions become questionnaires for the team. Whenever with-code uses external or additional repository path(s), ask for the local path(s) and check whether each path is readable right now: if readable, confirm the detected code signals; if outside the current session sandbox or not yet mounted, keep the confirmed path(s) for docs/agents/refinement.md and state how to grant read-only access before running workflows (adding the folder(s) to a Multi-root workspace in Antigravity, VS Code, or Cursor, or launching Claude Code with claude --add-dir <path>).

    Language. Recommend based on existing files or en-US by default: en-US or pt-BR. Explain in one line: defines the language of examination, question rounds, narratives, Gherkin criteria, and verdicts.

    Work items. Do User Stories originate in an external tracker? If yes, what is the base URL and field convention (for example Azure: <ID> for Azure DevOps work items). Every US originates there and each artifact carries this ID line.

    Product stage and context. Is this a new (greenfield) product or an ongoing (brownfield) product already in production or advanced development? Recommend based on explored code, CONTEXT.md, docs/product-overview.md, or documents in docs/:

    • New product: if CONTEXT.md does not exist yet, no action is needed now: the first examination that resolves a domain term will create it.
    • Ongoing product: explain that docs/reference/ is where the team places existing product manuals, business rules, specifications, or exported backlog stories so the agent does not enter story refinement blind, and ask whether existing documentation will be placed in docs/reference/, in another path under docs/ to list in ## Fact sources, or whether none exists yet. Note that running /bootstrap-context after setup reads those documents (and source code in with-code mode, or conducts a short onboarding examination when neither exists) to build CONTEXT.md and docs/product-overview.md before the first /refine-story.

    Completed when: mode, any with-code codebase path(s), language, work item convention, base URL, and product stage (including any reference documentation path for ongoing products) are confirmed.

  3. Confirm and write. Present the draft of docs/agents/refinement.md based on refinement.md (always writing Mode: with-code or Mode: without-code, migrating legacy com-codigo or sem-codigo in place when reconfiguring, listing any custom reference path under ## Fact sources, and filling the source code in clause of ## Fact sources per refinement.md, keeping . alongside any sibling repository paths when the current workspace also holds source code and omitting that clause in without-code mode), docs/agents/issue-tracker.md based on issue-tracker.md (only when it does not already exist), and the ## Refinement block for the agent instructions file. Allow adjustments. Then write the files, and when confirmed for an ongoing product with reference documentation, create docs/reference/ (or the confirmed custom reference path, with an empty .gitkeep when the folder does not yet exist) so the team sees immediately where to drop reference files. For the instructions file: if CLAUDE.md exists and imports AGENTS.md (@AGENTS.md), edit AGENTS.md; otherwise edit CLAUDE.md if it exists; otherwise AGENTS.md if it exists; if neither exists, ask which to create. If the ## Refinement block already exists, update it in place without modifying the rest of the file.

    The block:

    ## Refinement
    
    Socratic refinement workflows: run `/which-flow` to select a workflow. Mode and conventions live in `docs/agents/refinement.md`; tracker conventions in `docs/agents/issue-tracker.md`. Artifacts live in `.scratch/<slug>/` with `Status:` and work item ID.

    Completed when: refinement.md, the tracker file (new or preserved), the confirmed reference folder when ongoing with documentation, and the instruction block exist with confirmed content.

  4. Conclude. Confirm the workspace is configured, list workflows that read docs/agents/refinement.md, remind the user to grant read-only session access first if any configured codebase path was unreadable in step 2, and direct the user to the next step: for ongoing products without baseline context, drop existing manuals, rules, or exported stories into docs/reference/ (when confirmed) and run /bootstrap-context before /refine-story; for greenfield products, run /which-flow. Note that files can be edited manually later; re-running this skill is only required to switch modes, codebase paths, languages, or tracker conventions (already-configured workspaces carrying legacy com-codigo or sem-codigo continue to work without needing to re-run it). Completed when: workspace setup summary, workflows reading docs/agents/refinement.md, any pending codebase-access reminder from step 2, and the stage-appropriate next command (/bootstrap-context or /which-flow) are stated.

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.