Phase D Batch 01 Notes — Spec-Driven Secondary (Single-Author SDD Methodologies)

Batch roster

URL resolution notes

• autospec: URL not given directly. Resolved via gh api /repos/hesreallyhim/awesome-claude-code/issues/391 → https://github.com/ariel-frischer/autospec
• specorator-workflow: URL not given directly. Resolved via gh search repos "specorator" --limit 10 → Luis85/agentic-workflow (description: "The Specorator Workflow bundled in a Claude Plugin")
• All other URLs were given directly

Tier C stubs

None. All 10 frameworks were located and received full 11-file treatment.

spec-driver-davidlee relationship note

spec-driver-davidlee is the same repository as the seed spec-driver (davidlee/spec-driver). Set canonical: false in METRICS.yaml. The Phase D analysis adds v3 rubric fields absent from the seed record. Key additions: TUI (spec-driver tui) with live agent session follow mode, aggressive SessionStart hook via hookSpecificOutput injection, PostToolUse hook (artifact_event.py, async) enabling TUI follow mode, TOML-based workflow configuration.

Intra-batch patterns

1. Dominant pattern: "CLAUDE.md + spec files, no CLI"

Six of ten frameworks (094459-sdd-workshop, spec-based-claude-code, spec-first-nlatuan, spec-first-sunrain, spec-driven-fredantb, brian-disciplined-ai) follow the minimal distribution pattern: copy documentation files into project, no installable primitives. This is Archetype 4 (Markdown scaffold, zero Claude Code primitives).

2. Spectrum of automation

The batch spans the full automation spectrum:

• Zero automation: brian-disciplined-ai (no scripts, no hooks, no CLI)
• Script-only: spec-first-sunrain (51 shell scripts, no hooks)
• Hooks + skills: spec-driver-davidlee (2 hooks + 24 skills + TUI)
• Full platform: specorator-workflow (18 commands + 40 skills + 36 agents + MCP + Obsidian companion)

3. Cybernetics-inspired adaptive control

zhu1090093659-sdd is unique in applying control theory to AI behavioral management: 20%/40%/60% drift thresholds trigger progressively stronger behavioral corrections. No other framework in the corpus uses quantitative thresholds for AI behavioral correction.

4. GitHub-native tracking split

The batch splits on GitHub issue tracking:

• GitHub-native (use Issues/Projects as primary task tracker): zhu1090093659-sdd (GITHUB_FULL/GITHUB_STANDARD/LOCAL_ONLY modes), specorator-workflow (GitHub MCP integration)
• File-native (use local Markdown for task tracking): all others

5. Single-author web-browser-AI targeting

brian-disciplined-ai stands out as the only framework in the entire corpus explicitly designed for web browser AI (ChatGPT, Claude web, Gemini) as a primary use case — not just as a secondary target. The dual Markdown/PAG format exists entirely to serve this goal.

6. Scope inflation pattern

specorator-workflow is the most scope-ambitious framework in the batch — attempting to cover the entire software product lifecycle including Sales, Portfolio management, and Roadmap planning, not just the development workflow. This is significantly broader than any seed framework.

Most interesting find

zhu1090093659-sdd's cybernetics-inspired drift detection (866 stars).

The adaptive control loop using 20%/40%/60% threshold-based behavioral correction is conceptually novel — borrowing from control theory to treat AI behavioral drift as a systems control problem rather than a prompting problem. The high star count (866, highest in the batch) suggests this resonated with a broad audience despite being a single-author Chinese project. The three-mode GitHub integration (GITHUB_FULL/GITHUB_STANDARD/LOCAL_ONLY) is also practically thoughtful — recognizing that not all users want full GitHub automation.

Runner-up: brian-disciplined-ai's PAG (Pattern Abstract Grammar). A custom keyword grammar for instruction precision, designed for both web browsers and CLI agents simultaneously, is architecturally unique in the corpus.

Seed coverage

All 10 frameworks placed against seeds:

Cross-references to other batches

• spec-driver-davidlee: Direct relationship to the spec-driver seed (same repository). Any batch that references spec-driver should cross-reference this entry for v3 rubric additions.
• specorator-workflow: The Obsidian companion (specorator-obsidian-mcp) is a separate repository — if it appears in a future Obsidian/MCP-focused batch, cross-reference here.
• zhu1090093659-sdd: The cybernetics/adaptive control pattern may reappear in enterprise governance batches (Batch 28) — cross-reference if other frameworks use threshold-based behavioral correction.
• brian-disciplined-ai: PAG syntax is unique to this project. CC BY-SA 4.0 license is the only attribution-required license seen in Phase D so far — flag for legal/license comparative analysis.

Component primitive counts (batch summary)