agentleFS
gufrancogufranco/claude-engineering-rules/skills/zoom-out/SKILL.md

zoom-out

Skill3 starsChanged 31 days agoMITView on GitHub

Step back from a narrow code section and place it in its system context. Reads the file, traces callers and callees, identifies the module boundary, names the architectural role, and surfaces hidden coupling. Use when user says "zoom out", "what is this", "where does this fit", "give me context on this code", "what calls this", or feels lost in unfamiliar code. For line-by-line code review, use /review; for debugging, use /investigate; for onboarding to a whole repo, use /onboard.

114 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use to-prd, to-issues 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.
  • GoogleSkill

    to-prd

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

    Also in 60 projects that use this

  • GoogleSkill

    to-issues

    Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices.

    Also in 49 projects that use this

  • simo97Skill

    request-refactor-plan

    Use this skill when user wants to create a refactor plan.

    Also in 16 projects that use this

  • simo97Skill

    design-an-interface

    Generate multiple radically different interface designs for a module using parallel sub-agents. Use when user wants to design an API, explore interface options, compare module shapes, or mentions "design it twice".

    Also in 16 projects that use this

  • kennethkeimSkill

    ubiquitous-language

    Extract a DDD-style ubiquitous language glossary from the current conversation, flagging ambiguities and proposing canonical terms. Saves to UBIQUITOUS_LANGUAGE.md. Use when user wants to define domain terms, build a glossary, harden terminology, create a ubiquitous language, or mentions "domain model" or "DDD".

    Also in 15 projects that use this

  • Code-ShockSkill

    git-guardrails

    Set up git guardrails — a Claude Code PreToolUse hook that intercepts and blocks dangerous git commands such as force-push and hard reset before they run. Use to install git safety hooks or when the user runs /git-guardrails.

    Also in 3 projects that use this

  • +3 more from simo97
  • +1 more from google

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 5 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 24 edited versions we read keep these.

  • 5

    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:

    Read the full file at <path>.

    line 31
  2. Very specific:

    Stop at depth 1 unless --depth is higher.

    line 43
  3. Distinctive:

    If the target spans more than one role, that is itself a finding.

    line 74
  4. Distinctive:

    If the role does not match the table, say so explicitly.

    line 125
  5. Very specific:

    Stop after the report. Do not propose refactors.

    line 126
  6. Distinctive:

    Stop and run the caller grep when the Callers list is empty and the target has exports.

    line 137
  7. Very specific:

    Stop and cite the location when a claim in the report carries no file:line.

    line 138
  8. Distinctive:

    Stop at the report when the next step you are drafting is an edit or a refactor proposal.

    line 139
  9. Very specific:

    Stop and narrow --depth when the traced file count passes what fits in one report.

    line 140
  10. Very specific:

    Every caller, callee, and hidden-coupling line cites file:line.

    line 145
  11. Distinctive:

    No file was modified during the run.

    line 147

“Common” is measured: kept by at least 6 in 10 of the 24 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

Overview

Perspective shift. The user is staring at a function, file, or module and cannot see the system around it. This skill widens the lens until the role of the code is clear, then stops.

When to Use

  • The user says "zoom out", "what is this", "where does this fit", "give me context on this code", or "what calls this".
  • The user names one file or one line and cannot say which module owns it or who calls it.

For line-by-line review, use /review; for a bug hunt, use /investigate; for whole-repository orientation, use /onboard.

Arguments

  • <path> or <path>:<line>: the file or location to zoom out from. Required.
  • --depth <n>: how many hops outward to trace. Default 2, max 4. Higher depth burns context fast.

Process

1. Read the target

Read the full file at <path>. If a line number is given, read 30 lines around it. Identify:

  • The primary export or symbol at the location.
  • Other exports in the same file.
  • Imports from inside the project versus from third-party packages.

2. Trace inward (callees)

For each function the target calls, locate its definition. Note:

  • Which module it lives in.
  • Whether it is pure or has side effects such as I/O, network, database, or global state.
  • Whether it is project code or library code.

Stop at depth 1 unless --depth is higher.

3. Trace outward (callers)

Grep the project for usages of the target's exports. For each caller:

  • Which module imports it.
  • The shape of the call site, one-shot, loop, retry, conditional.
  • Whether the caller is a route handler, a job, a CLI entry, a test, or another internal module.

4. Identify the module boundary

Group the files touched in steps 2 and 3 by directory and by responsibility. State:

  • Which module the target belongs to.
  • Which neighboring modules it talks to.
  • Whether the boundary looks intentional or accidental such as cross-cutting imports, circular references, or leaky abstractions.

5. Name the architectural role

Pick the closest match from this list. Add a one-line justification.

RoleSignal
Domain logicPure functions, no I/O, expresses business rules
Application serviceOrchestrates domain calls and infrastructure, no business rules
AdapterTranslates between an external system and the domain
Controller / route handlerSits at the HTTP, queue, or CLI boundary
RepositoryPersists or retrieves domain objects
UtilityStateless helper used across modules
ConfigurationLoads, validates, or exposes config
Entry pointWires the system at startup

If the target spans more than one role, that is itself a finding. Surface it.

6. Surface hidden coupling

List anything that would surprise a new contributor:

  • Side effects at module load time. See the "No side effects at module level" bullet under "Fundamentals" in standards/code-style.md.
  • Shared mutable state.
  • Implicit ordering requirements between callers.
  • Logic that depends on environment variables read deep in the call stack.
  • Type assertions that hide a contract violation.

7. Output

Produce a single report. No file edits. Format:

# Zoom-out: <path>

## What it is
One paragraph. Primary purpose, surface area, dependencies.

## Role
<role from the table>. Justification: <one line>.

## Callers (depth N)
- file:line - <one line>
- file:line - <one line>

## Callees (depth N)
- file:line - <one line>
- file:line - <one line>

## Module boundary
- Lives in: <module>
- Talks to: <neighbor modules>
- Boundary health: clean / leaky / circular / cross-cutting

## Hidden coupling
- <finding> at file:line
- <finding> at file:line

## Suggested next reads
- file:line - reason
- file:line - reason

Rules

  • This skill never edits code. It produces a report only.
  • Cap depth at 4. Beyond that, the report stops being a zoom-out and becomes a survey.
  • Cite every claim with file:line. No abstract statements about "the architecture".
  • If the role does not match the table, say so explicitly. Do not invent a category.
  • Stop after the report. Do not propose refactors. Use /refactor for that.

Common Rationalizations

  • "The file name says what it does": a name is a claim. The callers in step 3 show the role the code actually plays.
  • "Depth 4 gives a fuller picture": past depth 4 the report turns into a survey, and the role of the target gets lost in it.
  • "It is close enough to Application service": a near match hides the finding. Say the role does not fit the table and describe what it does.
  • "The coupling is obvious, no need to cite it": every hidden-coupling item carries a file:line so the reader can check it in one step.

Red Flags

  • Stop and run the caller grep when the Callers list is empty and the target has exports.
  • Stop and cite the location when a claim in the report carries no file:line.
  • Stop at the report when the next step you are drafting is an edit or a refactor proposal.
  • Stop and narrow --depth when the traced file count passes what fits in one report.

Verification

  • The report has every heading from the step 7 template.
  • Every caller, callee, and hidden-coupling line cites file:line.
  • The role names one table entry with a justification, or says plainly that none fits.
  • No file was modified during the run.

Related skills

  • /onboard for whole-repo orientation.
  • /explain for line-level explanation of a block.
  • /refactor after the zoom-out reveals a structural problem.
  • /investigate when the goal is finding a bug, not understanding the structure.

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.