| name | methodology |
| description | AI-First development methodology: spec-driven pipeline, project structure,
skills/agents ecosystem, quality gates.
Use when: "изучи методологию", "изучи глобальную папку", "как работает методология",
"how does the methodology work", "explain the workflow"
For infrastructure tasks, use infrastructure-setup or deploy-pipeline skills.
|
AI-First Development Methodology
What Is This
A structured development approach for AI agents. Every feature goes through a pipeline: idea → spec → architecture → tasks → implementation → documentation update. Each stage has automated validators and quality gates. QA and deploy are regular tasks in the tech-spec, not separate pipeline steps.
Core problems it solves:
- Context loss between sessions — distributed knowledge base persists across sessions
- Quality without human review — automated validators at every stage
- Scope creep — specs approved before coding starts
- Outdated agent knowledge — Context7 MCP fetches current library docs
Development Pipeline
The full path from idea to production. Each step has a command, a skill behind it, and validators:
| Step | Command | Output | Skill |
|---|
| 1. User Spec | /new-user-spec | work/{feature}/user-spec.md (approved) | user-spec-planning |
| 2. Tech Spec | /new-tech-spec | work/{feature}/tech-spec.md (approved) | tech-spec-planning |
| 3. Task Decomposition | /decompose-tech-spec | work/{feature}/tasks/*.md (validated) | task-decomposition |
| 4. Implementation | /do-task or /do-feature | committed code + reviews | from task file / feature-execution |
| 5. Done | /done | PK updated, feature archived | documentation-writing |
Step 4 mode choice:
/do-task — single task, manual control, debugging, iterating on one piece./do-feature — multiple tasks ready, standard feature work, parallel execution via agent teams.
To run or explain any step, follow its process in pipeline.md — per-step validators, commit points, iteration limits, wave mechanics, escalation rules, plus cross-cutting principles (commit strategy, validation counts, checkpoint recovery, Context7, just-in-time context).
What Do You Need?
Run a pipeline step / explain how a stage works?
→ read pipeline.md — steps 1–5 in detail + key principles
Find where docs, specs, tasks live (Project Knowledge, work/{feature}/, global ~/.claude/)?
→ read project-structure.md
Look up a skill, agent, or slash command by name or category?
→ read ecosystem.md — planning/execution/quality/meta skills, validators, reviewers, commands
Apply the cross-skill behavioural rules in detail (formats, failure modes)?
→ read operating-behaviors.md
Load the smallest set of references that fits the task.
Core Operating Behaviors (apply across all skills)
Every agent in this methodology follows these, regardless of which skill is active. The pipeline (specs, validators, reviewers) catches outputs; these behaviours catch inputs — silent failure modes that bypass validators:
- Surface assumptions — before non-trivial work, list assumptions explicitly so the user can correct them.
- Manage confusion actively — on inconsistency: stop, name the confusion, present the tradeoff, wait for resolution.
- Push back when warranted — point out clear problems, quantify the downside, propose an alternative.
- Enforce simplicity — prefer boring, obvious solutions; abstractions earn their complexity.
- Scope discipline — touch only what the task requires; record side-observations as
NOTICED BUT NOT TOUCHING:.
- Verify, don't assume — a task is complete only when verification evidence exists (tests, build output, smoke check).
When applying these during work, follow the formats and rules in operating-behaviors.md — assumption-listing format, which skills ritualise each behaviour, and the 8 failure modes they catch.
Workflow Quick Start
New project:
/init-project → /init-project-knowledge (interview + fill all docs) → start features
New feature:
/new-user-spec → /new-tech-spec → /decompose-tech-spec → /do-feature or /do-task → /done
Ad-hoc coding (no spec):
/write-code
To understand how a specific skill works internally, read its SKILL.md directly.
