{
  "markdown": "# gemini-redteam\nRed team SKILL\n\n\nA multi agentiic workflow that injects a professional offensive security persona and provides red team workflows.\n\n## Install\n\n### multi agent skills\n\n```bash\n# From GitHub\nnpx skills add Baba01hacker666/gemini-redteam\n```\n\nThen restart your multi agent skills session.\n\n### Google Antigravity\n\n```bash\nagy install https://github.com/Baba01hacker666/gemini-redteam\n```\n\n## What it does\n\n- Injects a **security research system context** into multi agent skills sessions through `GEMINI.md`\n- Provides **6 multi agent skills slash commands** for common red team workflows\n- Bundles **4 multi agent skills sub-agents** for CMS fingerprinting, source analysis, finding verification, and final report writing\n- Bundles **6 skills** for multi agent skills-compatible and Antigravity workflows\n- Provides an **Antigravity plugin manifest** and red team rule under `plugin.json` and `rules/`\n\n## Design goals\n\n- **Operational outputs**: command-first responses with concrete next actions\n- **Prompt robustness**: templates request clarifications when critical context is missing\n- **Evidence-aware writing**: report and exploit flows separate facts, assumptions, and hypotheses\n- **Consistent structure**: every command now pushes for validation checkpoints and fallback paths\n- **Anti-hallucination verification**: analyzer outputs must be checked by an independent verifier before final reporting\n- **Evidence standardization**: verified findings require affected state proof, exploitability or impact proof, negative controls, reproduction metadata, and retest steps\n- **Report artifact**: workflows end by creating or updating `report.md` with steps tried, evidence, verified results, and gaps\n\n## Commands\n\n| Command | Usage |\n|---|---|\n| `/rt:recon <target>` | Structured recon plan for a target type/scope |\n| `/rt:exploit <vuln/context>` | Exploit chain analysis and PoC |\n| `/rt:ctf <challenge>` | CTF solver — enum to flag |\n| `/rt:cms <cms/target>` | CMS-specific assessment order, checks, and verification |\n| `/rt:evade <payload/context>` | AV/EDR evasion strategy |\n| `/rt:report <finding>` | Professional pentest finding write-up |\n\n## Bundled multi agent skills agents\n\nmulti agent skills loads extension sub-agents from the `agents/` directory. You can ask the main session to delegate automatically, or force a specific agent with `@agent-name`.\n\n| Agent | Use when | Output |\n|---|---|---|\n| `@redteam-cms-fingerprint` | You need CMS/platform identification and component inventory before testing | CMS confidence, components, prioritized checks, analyzer/verifier handoff |\n| `@redteam-source-code-analyzer` | You need source-code review for routes, controllers, plugins, modules, themes, sinks, and candidate vulnerabilities | Evidence-backed candidate findings (`SC-001`, `SC-002`, ...), commands/files reviewed, verifier handoff |\n| `@redteam-finding-verifier` | Another agent reported possible findings and you need to confirm they are real | Confirmed / partially confirmed / unproven / rejected verdicts with hallucination checks |\n| `@redteam-report-writer` | The workflow is complete and a durable artifact is needed | Creates or updates `report.md` with full steps, attempts, observations, verified findings, rejected claims, remediation, and retest steps |\n\nRecommended agent order for CMS/source reviews:\n\n```text\n@redteam-cms-fingerprint → @redteam-source-code-analyzer → @redteam-finding-verifier → @redteam-report-writer\n```\n\nThe final report writer must not turn analyzer hypotheses into findings unless the verifier confirms them. If no issue is verified, `report.md` must say no verified vulnerabilities were confirmed.\n\n## Bundled skills\n\nmulti agent skills loads extension skills from the `skills/` directory. The broad `redteam` skill remains available, and focused skills provide lighter workflow-specific context.\n\n| Skill | Use when |\n|---|---|\n| `redteam` | General offensive-security research, CMS testing, exploit analysis, CTFs, and reporting |\n| `redteam-recon` | Building recon plans, enumerating attack surface, and prioritizing pivots |\n| `redteam-cms` | Fingerprinting and assessing CMS platforms, plugins, modules, themes, and extensions |\n| `redteam-source-audit` | Reviewing repository source code and producing verifier-ready candidate findings |\n| `redteam-exploit-validation` | Safely validating exploitability with proof ladders, controls, cleanup, and telemetry notes |\n| `redteam-reporting` | Creating evidence-backed findings and `report.md` artifacts from verified work |\n\n## Examples\n\n```\n/rt:recon Laravel application exposed on Shodan, unknown version\n/rt:cms WordPress multisite with WooCommerce, authenticated editor account\n/rt:exploit CVE-2025-54236 Magento unauthenticated RCE\n/rt:ctf PHP web challenge with a file upload endpoint and source provided\n/rt:evade Cobalt Strike beacon, x64, Windows 11 with Defender + CrowdStrike\n/rt:report SQL injection in /api/users?id= leading to full DB dump\n```\n\n## Structure\n\n```\ngemini-redteam/\n├── gemini-extension.json     # multi agent skills extension manifest\n├── plugin.json               # Antigravity plugin marker\n├── GEMINI.md                 # multi agent skills system context\n├── commands/\n│   └── rt/\n│       ├── recon.toml\n│       ├── exploit.toml\n│       ├── ctf.toml\n│       ├── cms.toml\n│       ├── evade.toml\n│       └── report.toml\n├── agents/\n│   ├── redteam-cms-fingerprint.md\n│   ├── redteam-source-code-analyzer.md\n│   ├── redteam-finding-verifier.md\n│   └── redteam-report-writer.md\n├── docs/\n│   └── report-template.md\n├── rules/\n│   └── redteam.md            # Antigravity red team operating rule\n└── skills/\n    ├── redteam/\n    │   └── SKILL.md          # Broad offensive-security skill\n    ├── redteam-recon/\n    │   └── SKILL.md          # Recon and attack-surface skill\n    ├── redteam-cms/\n    │   └── SKILL.md          # CMS assessment skill\n    ├── redteam-source-audit/\n    │   └── SKILL.md          # Source-code review skill\n    ├── redteam-exploit-validation/\n    │   └── SKILL.md          # Safe exploit validation skill\n    └── redteam-reporting/\n        └── SKILL.md          # Evidence-backed reporting skill\n```\n\n## Notes\n\n- multi agent skills context is injected globally through `GEMINI.md` when the extension is active.\n- Antigravity loads `plugin.json`, `skills/`, and `rules/` when this repo is placed in a supported plugin directory.\n- multi agent skills commands conflict-resolve with prefix: if `/recon` exists elsewhere, these become `/gemini-redteam:rt:recon`.\n- multi agent skills sub-agents in `agents/` are multi agent skills-specific; Antigravity support currently uses skills and rules.\n- Skills are token-efficient — only activated when relevant.\n\n## Recent prompt quality improvements\n\n- Added explicit handling for ambiguous input (short clarifying questions before execution).\n- Added validation checkpoints so users can confirm each step succeeded.\n- Added CMS-specific workflows for WordPress, Drupal, Joomla, Magento/Adobe Commerce, Shopify, Ghost, Strapi, Umbraco, Sitecore, TYPO3, PrestaShop, and OpenCart.\n- Added multi agent skills sub-agents for CMS fingerprinting, source code analysis, finding verification, and `report.md` generation.\n- Added focused skills for recon, CMS assessment, source audit, exploit validation, and evidence-backed reporting.\n- Added Google Antigravity plugin support with `plugin.json` and `rules/redteam.md`.\n- Added a shared evidence standard for affected-state proof, exploitability/impact proof, negative controls, reproduction metadata, and remediation retests.\n- Added detection/telemetry considerations in exploit and evasion outputs.\n- Added assumptions/limitations guidance in reporting output when evidence is incomplete.\n",
  "bytes": 7766,
  "sha": "fd41453121cbb4f4bcb5086e412d527711acd93a96cd858bf819ea58ed3fb9fc",
  "repo_slug": "baba01hacker666/gemini-redteam",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_baba01hacker666_gemini_redteam_554d8c8a/readme"
}