agentleFS
omspradippatilomspradippatil/Claude-Skills/.claude/skills/gsd-core/SKILL.md

gsd-core

Skill3 starsChanged 10 days agoMITView on GitHub

Activate this skill for complex features, multi-stage milestones, major refactoring, and long-running development tasks requiring disciplined phase loops (Discuss, Research, Plan, Execute, Verify) and persistent project state artifacts. Do not use for simple bug fixes or routine edits.

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If subagents are supported, delegate deep codebase scans to isolated subagent contexts.

    line 56
  2. Distinctive:

    Check actual task acceptance criteria directly.

    line 73
  3. Very specific:

    Durable Milestones: Write architectural decisions, active phase, and completed checklist items to a tracking markdown file in the project …

    line 81
  4. Distinctive:

    Context Resets: When conversation history becomes large, summarize progress into the persistent state file so a fresh turn can resume …

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

GSD Core — Spec-Driven Complex Development Skill

Official upstream: open-gsd/gsd-core (MIT License)


Philosophy & Core Objective

"Git. Ship. Done."

Complex engineering tasks fail when agents lose context ("context rot"), attempt to solve too many things in a single turn, skip architectural alignment, or claim completion without verification.

GSD Core enforces a structured, phase-bounded development lifecycle designed to keep context focused, code thoroughly grounded, and progress auditable across sessions.


When to Activate

  • Building non-trivial multi-file features or services.
  • Major system refactoring or architectural migrations.
  • Greenfield project initialization or multi-milestone roadmaps.
  • Tasks spanning multiple engineering domains (e.g., fullstack API + UI + Database + Security).

When NOT to Activate

  • Single-line edits, typo fixes, or small bug fixes (use lightweight execution under policies/engineering.md).
  • Simple documentation edits or routine shell commands.
  • Standalone research questions.

The 5-Phase Lifecycle

┌───────────┐     ┌───────────┐     ┌───────────┐     ┌───────────┐     ┌───────────┐
│  Discuss  │ ──► │ Research  │ ──► │   Plan    │ ──► │  Execute  │ ──► │  Verify   │
└───────────┘     └───────────┘     └───────────┘     └───────────┘     └───────────┘

Phase 1: Discuss

  • Align on the user's core intent, functional requirements, and constraints.
  • Define explicit acceptance criteria before touching code.
  • Clarify edge cases, performance targets, and security expectations.
  • Agree on what is in scope and what is explicitly out of scope.

Phase 2: Research

  • Inspect the codebase to understand existing conventions, data flows, and callers.
  • Identify external dependencies, APIs, and potential breaking changes.
  • If subagents are supported, delegate deep codebase scans to isolated subagent contexts.
  • Resolve technical unknowns before writing implementation plans.

Phase 3: Plan

  • Break the milestone into small, sequential, testable tasks.
  • Establish dependency order: database ➔ backend API ➔ frontend UI ➔ integration tests.
  • Define the exact write scope for each task.
  • Persist plan and decisions in durable project artifacts (PROJECT.md, ROADMAP.md, or task state) so work survives context resets.

Phase 4: Execute

  • Implement one bounded task at a time.
  • Ground before mutating: read existing files and tests before modifying.
  • Adhere strictly to the declared write scope; do not wander into unrelated files.
  • Apply Ponytail minimal engineering: write the smallest correct solution.

Phase 5: Verify

  • Require fresh evidence: run validators, type checks, unit tests, and integration tests after final mutations.
  • Check actual task acceptance criteria directly.
  • Report honest findings: distinguish executed evidence from assumptions or suggestions.

State Persistence & Context Management

To prevent context rot during large tasks:

  1. Durable Milestones: Write architectural decisions, active phase, and completed checklist items to a tracking markdown file in the project (e.g., .claude-skills/state/task-plan.md or ROADMAP.md).
  2. Context Resets: When conversation history becomes large, summarize progress into the persistent state file so a fresh turn can resume cleanly.
  3. Subagent Delegation: Where supported (e.g. Antigravity subagents), offload parallel investigations or bounded implementation tasks to child agents. Retain the main session for orchestration, dependency tracking, and final verification.

Integration with Repository Policies

GSD Core orchestrates the workflow lifecycle, while repository policies govern technical rules:

  • Gravitas Core: Supplies the six reliability invariants and completion evidence requirements.
  • Policies (Security, SEO, Frontend, Testing): Authoritative rules for each specific domain.
  • Ponytail: Guides minimal, unbloated code implementation within Phase 4.
  • Caveman: Keeps progress reports and execution logs concise and focused.

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.

Reviews can't be read right now.

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.