adversarial-spec

Brief → structured spec. Two-role adversarial pipeline that takes a product brief, specification request, or feature idea and produces a spec.md with YAML frontmatter, numbered requirements, and acceptance criteria.

For Hermes Agent, Claude Code, Codex, or any LLM CLI.

How it works

PHASE 1 ──→ WRITE     (spec-writer turns brief into structured spec.md)
PHASE 2 ──→ CHALLENGE (spec-challenger attacks for gaps, contradictions, untestable criteria)
PHASE 3 ──→ REVISE    (spec-writer amends per findings)
PHASE 4 ──→ VERIFY    (spec-challenger checks findings resolved)
MERGE  ──→ squash-merge (APPROVED) or [REJECTED] commit

Output format

---
name: "feature-name"
version: "1.0"
status: "draft"
targets:
  - file: path/to/file
    description: "What this file must do"
---

## Requirements
- R1: …
- R2: …

## Acceptance criteria
- AC1 (R1): …
- AC2 (R2): …

Specs are consumed by adversarial-plan for planning and adversarial-code-loop for implementation.

Comparison

Feature	adversarial-spec	zscole/adversarial-spec
Structured frontmatter	✅ YAML + targets + requirements	❌ Free-form
Acceptance criteria	✅ Per-requirement ACs	❌
Git-native pipeline	✅ Branch-per-spec, squash-merge	❌ Single file
plan/code-loop integration	✅ Direct feed to plan + loop	❌ Standalone

Quick start

python3 scripts/adversarial_spec.py \
  --brief "A feature to …" \
  --dev-cmd "pi --provider zai --model glm-5.2" \
  --review-cmd "pi --provider deepseek --model deepseek-v4-pro"

Dependencies

• Python ≥ 3.11
• Git ≥ 2.5
• Two LLM CLIs (spec-writer + spec-challenger)

Uses adversarial-common as the shared engine.

License

MIT