Manage long-running AI agent development projects with incremental progress, scoped features, and verification#2
Open
sunfmin wants to merge 17 commits intotheplant:mainfrom
Open
Manage long-running AI agent development projects with incremental progress, scoped features, and verification#2sunfmin wants to merge 17 commits intotheplant:mainfrom
sunfmin wants to merge 17 commits intotheplant:mainfrom
Conversation
Rename iterative-web-dev to iterative-dev with support for web, API, CLI, library, data pipeline, and mobile projects. The core autonomous loop is unchanged — only verification strategy and applicable standards swap based on a "type" field in feature_list.json. - Restructure references/ into core/, web/, and verification/ subdirs - Add 6 type-specific verification strategies (web, api, cli, library, data, mobile) - Update feature_list.json format with top-level type field and type-specific categories - Make subagent template, post-verification checks, and decision guidelines type-aware - Expand session-handoff and init-script docs to cover Go, Python, Rust, Node.js - Add MIT license - Bump to v2.0.0 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Show 5 cases: manual spec, agent-generated spec, scope switching, compliance audits, and continuing multi-session projects. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
progress.txt is scope-specific, so it belongs inside specs/{scope}/
alongside spec.md and feature_list.json, symlinked to the project root.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Generalize to any project type
…fication Subagents could not find reference docs because paths were relative to the skill install directory, not the project. This resolves all reference paths to absolute paths in subagent prompts. Also inlines screenshot capture instructions into the subagent template and adds a non-negotiable screenshot gate in the parent agent's post-verification flow. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Features must include their own test/verification steps instead of deferring testing to separate features. This prevents false progress where features appear "done" but are unverified. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
sunfmin
changed the title
Web/mobile UI features must now include explicit screenshot capture, Playwright verification, and visual review steps. This ensures visual quality is verified within each feature rather than deferred to a separate gate. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…prefix bugs When a web project connects its frontend to a real backend API, TypeScript compilation alone cannot catch two common infrastructure issues: 1. Route prefix mismatch — code generators (ogen, openapi-generator) register routes without the OpenAPI servers.url prefix, so the backend serves at /products instead of /api/v1/products 2. Missing CORS headers — browsers silently block cross-origin requests, causing the frontend to show loading spinners forever Changes: - web-verification.md: Added "Full-Stack Integration Smoke Test" section with step-by-step process, fail-fast criteria, and common root causes table - SKILL.md: Added integration smoke test gate to parent agent post-verification, added conditional instructions to subagent prompt for API-connecting features, added 3 new entries to Decision Making Guidelines table - init-script-template.md: Added CORS and route prefix verification steps to the web project template with warning messages Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… outcome-oriented - e2e-verification.md: reframe screenshots as secondary to interaction tests, add subagent reference note distinguishing it from web-verification.md - init-script-template.md: generalize CORS/route-prefix section to "cross-component connectivity" with example framing instead of hardcoded paths - ux-standards.md: replace Tailwind-specific classes (text-2xl, shadow-sm, p-1, etc.) with technology-agnostic descriptions - code-quality.md: replace data-testid assumption with project-type-appropriate stable test selectors - SKILL.md: project type detection now asks about user interaction (browser, terminal, import/call) instead of specific frameworks; init-scope screenshot emphasis reframed with outcome-proving tests as primary verification - feature-list-format.md: add outcome-oriented features rule, rewrite verification steps to prove user outcomes not component existence - web-verification.md: add interaction test emphasis as primary verification, add correct/wrong test examples, reorder parent post-verification checks Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ent phase
Move screenshots from global e2e/screenshots/ into per-scope directories
(specs/{scope}/screenshots/) so artifact provenance is clear. Add a
refinement phase after each feature that launches a dedicated subagent to
polish UX/visual design (divergent thinking, research) and code quality
(abstraction, testability, maintainability). Refinement analysis is
persisted to specs/{scope}/refinements/feature-{id}-refinement.md for
traceability across sessions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Full-stack domain features must implement backend and frontend together in one feature, not split by technology layer. This catches CORS, route prefix, and response format issues during development instead of at integration time. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Refinement subagents were being skipped because there was no blocking gate — unlike screenshots which have concrete "check exists → BLOCK" enforcement. Added REFINEMENT GATE with same enforcement pattern. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Extracted subagent prompt templates to references/templates/ and removed duplicated rules that were restated 3-4 times across sections. The core loop is now clear: implement → verify → refine → next. No concepts lost — all moved to reference files or condensed. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- tests/verify.sh — checks artifacts after skill runs:
- All features pass
- feat: commits exist for each feature
- refine: commits exist for each feature (catches the skipped-refinement bug)
- Refinement reports exist in specs/{scope}/refinements/
- Screenshots exist (web/mobile)
- Commit pattern: no consecutive feat: without refine: between them
- tests/smoke-test.md — minimal 2-feature test scope for quick validation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…eral artifacts Screenshots are project deliverables that should be preserved in the repo as evidence of feature verification. Removed all rm -rf commands targeting screenshot directories from init.sh templates and the e2e verification workflow. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Each refinement pass now creates a new file (e.g. feature-2-refinement-20260320-143052.md) instead of overwriting the previous report. This preserves the full history of reviews across multiple refinement passes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Manage long-running AI agent development projects with incremental progress, scoped features, and verification. Works with any project type — web, API, CLI, library, data pipeline, mobile. Use this skill when working on multi-session projects, implementing features incrementally, running tests, initializing project scopes, or continuing work from previous sessions. Triggers on phrases like "continue working", "pick up where I left off", "next feature", "run tests", "verify", "initialize scope", "switch scope", "feature list", "incremental progress", or any multi-session development workflow.