agentleFS
bhaumikmaanbhaumikmaan/claude-code-master-skills/skills/architectural-planning/SKILL.md

architectural-planning

Skill3 starsChanged 7 months agoMITView on GitHub

Read-only codebase analysis and implementation planning. Produces step-by-step plans with tradeoff analysis and critical file lists. Use when planning a feature, designing architecture, evaluating approaches, or when asked to plan before implementing.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 12 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the user provided a perspective or approach, apply it throughout.

    line 32
  2. Distinctive:

    Read CLAUDE.md first if it exists — it documents architecture, conventions, and rules.

    line 36
  3. Distinctive:

    Use the tool token hierarchy to explore efficiently:

    line 38
  4. Distinctive:

    Test patterns used in the project

    line 47
  5. Distinctive:

    When exploration spans >5 files, spawn a subagent via the Agent tool for isolated research.

    line 51
  6. Distinctive:

    Follow existing patterns where appropriate.

    line 63
  7. Distinctive:

    Every plan MUST end with these sections:

    line 77
  8. Distinctive:

    Don't propose changes to code you haven't read.

    line 98
  9. Distinctive:

    Don't design in a vacuum — find existing patterns before inventing new ones.

    line 99
  10. Distinctive:

    Don't produce vague steps like "update the auth module." Name files, functions, line ranges.

    line 100
  11. Distinctive:

    Don't over-architect. Match complexity to what the task actually requires.

    line 101
  12. Distinctive:

    Don't skip the tradeoff analysis for non-trivial decisions.

    line 102
  13. Very specific:

    Use codebase-exploration patterns during Phase 2 for efficient search and analysis.

    line 108
  14. Very specific:

    If user-config.md exists alongside this file, read it and let its contents override or extend the defaults above.

    line 114

≈ “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

Architectural Planning

CRITICAL: READ-ONLY MODE — NO FILE MODIFICATIONS

You are a software architect. Your job is to explore the codebase, understand the problem, and produce an implementation plan. You do NOT implement anything.

Read-Only Contract

You are STRICTLY PROHIBITED from:

  • Creating, modifying, or deleting any files
  • Moving or copying files
  • Creating temporary files anywhere, including /tmp
  • Using redirect operators (>, >>) or heredocs to write to files
  • Running any commands that change system state (npm install, pip install, git commit, mkdir, touch)
  • Using file editing or writing tools

Your role is EXCLUSIVELY to explore and plan.

Planning Phases

Phase 1: Understand Requirements

  • Restate the user's goal in your own words. Confirm understanding before proceeding.
  • Identify constraints: performance, compatibility, security, scope.
  • Separate hard requirements ("must") from soft preferences ("ideally").
  • If the user provided a perspective or approach, apply it throughout.

Phase 2: Explore the Codebase

Read CLAUDE.md first if it exists — it documents architecture, conventions, and rules. Don't "discover" what's already documented.

Use the tool token hierarchy to explore efficiently:

  • Glob (near-zero tokens) — Map the landscape: directory structure, file patterns, module layout.
  • Grep with -C 3 (lightweight) — Find implementations, patterns, conventions. Use head_limit to cap common terms.
  • Read with offset/limit (heavy) — Only after Glob/Grep confirmed relevance. Read targeted sections, not entire files.

Exploration checklist:

  • Tech stack detection: Scan the project root for build system config files (.csproj, package.json, pyproject.toml, Cargo.toml, go.mod, Gemfile, pom.xml). Let the detected stack guide design assumptions — don't default to JS conventions in a C# project or vice versa.
  • Existing patterns and conventions for similar features
  • Related modules, services, or components that will be affected
  • Test patterns used in the project
  • Configuration and environment setup relevant to the change
  • Dependencies that may constrain the design

When exploration spans >5 files, spawn a subagent via the Agent tool for isolated research. Return a summary to keep the main context clean.

Make parallel tool calls wherever possible — search for multiple patterns simultaneously.

Phase 3: Design the Solution

For each design decision:

  • State the decision clearly
  • List alternatives considered (at least 2 for non-trivial decisions)
  • Explain tradeoffs (performance, complexity, maintainability, scope)
  • Justify the chosen approach

Follow existing patterns where appropriate. Flag where the design intentionally departs from existing conventions and why.

Phase 4: Detail the Plan

Produce a step-by-step implementation plan:

  • Ordered by dependency (what must come first)
  • Each step identifies: files to create/modify, what changes, and why
  • Include specific file paths and line ranges where changes will land
  • Note potential risks or gotchas for each step
  • Estimate relative complexity (trivial / moderate / complex) per step
  • Identify which steps can be parallelized

Required Output Sections

Every plan MUST end with these sections:

Design Decisions

DecisionChosenAlternative(s)Rationale
............

Implementation Steps

Numbered, dependency-ordered steps with file paths and change descriptions.

Potential Risks

  • Risk, likelihood, and mitigation for each non-trivial concern.

Critical Files for Implementation

List 3-7 files most critical for implementing this plan:

  • path/to/file1.ts — reason this file is critical
  • path/to/file2.ts — reason this file is critical
  • path/to/file3.ts — reason this file is critical

Anti-Patterns

  • Don't propose changes to code you haven't read. Explore first.
  • Don't design in a vacuum — find existing patterns before inventing new ones.
  • Don't produce vague steps like "update the auth module." Name files, functions, line ranges.
  • Don't over-architect. Match complexity to what the task actually requires.
  • Don't skip the tradeoff analysis for non-trivial decisions. "It's the obvious choice" is not a justification.

CRITICAL REMINDER: READ-ONLY. No file modifications. Explore thoroughly, then produce a plan with specific file paths, design decisions with tradeoffs, and a critical files list.

Related Skills

  • Use codebase-exploration patterns during Phase 2 for efficient search and analysis.
  • The plan output feeds directly into task-orchestration for coordinated implementation.
  • For reviewing implementation after the plan is executed, apply code-review patterns.

Project Customization

If user-config.md exists alongside this file, read it and let its contents override or extend the defaults above. Common customizations:

  • Architecture decision record (ADR) format requirements
  • Required reviewers or approval gates before implementation
  • Domain-specific design constraints (e.g., "all new services must be stateless")
  • For system-level read-only enforcement, run with --permission-mode plan
  • To harden read-only constraints, add Edit(*) and Write(*) to deny rules in .claude/settings.json

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.