Skip to content

feat: add project-workflow skill — unified six-phase engineering life… - #326

Closed
zzx666-code wants to merge 1 commit into
addyosmani:mainfrom
zzx666-code:main
Closed

feat: add project-workflow skill — unified six-phase engineering life…#326
zzx666-code wants to merge 1 commit into
addyosmani:mainfrom
zzx666-code:main

Conversation

@zzx666-code

Copy link
Copy Markdown

…cycle orchestrator## Summary

Adds project-workflow, a meta-skill that orchestrates the entire agent-skills library
through a single, gate-controlled six-phase lifecycle (DEFINE → PLAN → BUILD → VERIFY
→ REVIEW → SHIP). Users can trigger it with a single sentence ("帮我写个项目",
"fix this bug", "add a feature"), and the agent automatically routes to the correct
sub-skills, enforces human checkpoints between phases, and produces verifiable evidence
at each gate.

Motivation

agent-skills has 20+ individual skills, but using them effectively requires knowing
which skill to invoke at which point. There was no single entry point that:

  • Routes a vague request to the right phase (e.g., bug reports skip DEFINE/PLAN
    and go directly to VERIFY's Prove-It pattern)
  • Enforces phase gates with explicit user confirmation before advancing
  • Auto-activates the correct sub-skills based on context (e.g., frontend-ui-engineering
    when UI work is detected, security-and-hardening when auth is involved)
  • Provides a shared code of conduct (three core operating principles + six bottom-line
    rules) that governs all sub-skill behavior

What's Included

  • skills/project-workflow/SKILL.md — the full skill definition (276 lines).
    Covers: routing table by confidence level, six phases with evidence checklists,
    10 auto-activating sub-skills by context, bilingual Chinese/English triggers,
    comparison table vs plain agent-skills, anti-rationalization tables, and
    verification checklist.
  • skills/project-workflow.zip — distribution package.

Structure of the Skill

Section Content
Overview Three core operating principles (Surface Assumptions, Manage Confusion Actively, Push Back When Warranted)
Six-Phase Lifecycle DEFINE → PLAN → BUILD → VERIFY → REVIEW → SHIP, each with sub-skill routing, auto-activation table, and evidence checklist
Bottom-Line Principles Six non-negotiable rules (Spec before code, Tests are proof, Thin vertical slices, etc.)
Usage Chinese and English trigger examples
How It Differs 7-dimension comparison vs using agent-skills directly
Red Flags & Verification Behavioral guardrails and exit criteria

How to Test

Load the skill into any agent that supports Markdown skill discovery. Trigger it with:

  • "帮我写个项目" — should route to DEFINE phase
  • "Fix this bug: the login page crashes on empty password" — should route to VERIFY
  • "Add a feature: task list filtering" — should route to DEFINE, write a spec

@addyosmani

Copy link
Copy Markdown
Owner

Going to close this one. A "unified six-phase lifecycle orchestrator" is what using-agent-skills already is, so this duplicates the meta-skill rather than adding a capability, and the branch has gone stale on top of that. The instinct to give people a single front door is right, it just happens to be the front door we already have. Appreciate the effort, and please keep contributing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants