adversarial-breaker
Adversarial verification — fixer→breaker chain. Independent offensive agent.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- adversarial-breaker
- triggers
- adversarial, breaker, !breaker, verify fix, romper, verificar fix, try to break, offensive verification
- changelog
- docs/ciclos/cycle28-20260815.md
- description
- Adversarial verification — fixer→breaker chain. Independent offensive agent.
- token_budget
- 3000
02 资源文件 5
- 参考文档references/attack-surface.md2.1 KB
- 参考文档references/breaker-briefing.md7.0 KB
- 参考文档references/profiles/node.md2.1 KB
- 参考文档references/profiles/powershell.md3.0 KB
- 参考文档references/profiles/python.md2.1 KB
03 SKILL.md 正文
When to Use
fixer done+ROJA/AMARILLA(auth/storage/API)+not config-only. Skip:VERDE·non-code·comments·docs. Token-save:diff<10L+config/lock/docs.
Protocol
1.Bundle:diff+files+fixer_claims+test_results+zone+pipeline_mode
2.Engram:mem_search("breaker/{target}")→## Past Attack Context
3.Zone:review-rules.jsonc. AMARILLA+auth→ROJA
4.Profiles: .agents/attack-surface.{project}.md→inject
5.Pre-reg:git stash→tests→pop. Break→PRE-BREAK
6.Dispatch:diff>50/ROJA→Parallel Break-7(3). Else→single
7.Launch:breaker-briefing+profile(s)+past+bundle
8.Parse:4-field. Parallel→merge. <3→FAIL→rerun
9.Calibrate:depth/relevance/coverage/specificity
10.Verdict:adjusted+PRE-BREAK
11.R2:FIX→round 2
12.Record:Engram+test cases
Verdicts
AllPASS→APPROVED|1-2minor→FIX|Critical→BLOCK|<3/malformed→FAIL|Timeout→ESCALATE
AB-{t}|R:{N}/2|A:{n}|SAFE/BROKEN|V:{v}
R2&Esc
new diff after fix; both diffs—R1+new. ##R2—Focus:{R1}. Max2. Escalate:Target/Rounds/Pipeline/Chain/Blocker/Rec
Integration
quality-gate(BEFORE)|triple-verify(AFTER)|judgment-day|subagent-isolation|external-auditor(BLOCK→confirm)|immune-system(repeated→fix)
Projects:.agents/attack-surface.{project}.md+.agents/breaker-profiles/{project}/
Anti-Patterns
Breaker trusts fixer·No attempts·3+r·Before QG·4R review·Happy path·Non-code/tiny diffs·No pre-reg·No test cases·Generic surface
docs/skills/adversarial-breaker/reference.md
Anti-Rationalization
| Rationalization | Red Flag | Verification |
|---|---|---|
| "el breaker ya pasó el profile scan, no hace falta re-verificar" | Reuse de profile sin fresh scan por cambio | Re-ejecutar attack-surface.{project}.md injection + mem_search breaker/{target} por cada diff |
| "breaker y fixer pueden ser el mismo agente, es más eficiente" | Mismo agente verifica su propio fix | Delegar breaker independiente vía subagent-isolation con bundle separado, nunca reutilizar fixer context |
| "clearance permanente tras un PASS" | Asumir aprobación vale para cambios futuros | Clearance es por-cambio: nuevo diff → nuevo ciclo breaker con verdict fresco AB-{t} |
Red Flags
- Doing work without checking output format → STOP, re-read skill
- Second occurrence of same rationalization → force RED zone
Verification
- Output matches skill ## Output contract + file:line citaton
- cross-ref-check.ps1 → SKILL.md OK
Refs
Cross-Refs: security-scanner | best-practices | judgment-day
04 同仓库更多 Skills
web-quality-audit
Comprehensive web audit: performance, a11y, SEO, responsive, animation, design tokens.
ui-engine
UI system — Grid/Flexbox/@layer/:has(), container queries, compositor-only animation, OKLCH tokens, component patterns
baseline-ui
Anti-slop UI — layout, typography, responsive, animation, tokens. Use for cleanup or polish.
skill-creator
Create new AI agent skills following Agent Skills spec -- auto-bootstrap unknown tasks, define structure, register in AGENTS.md
senior-engineer
Apply senior engineering competencies — system design, trade-off analysis, delegation, mentoring, and tech debt strategy
prompt-engineering
Design effective prompts using SPEARS framework — scope, principles, examples, assertions, refinements, with ReAct and multi-agent patterns