[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1173":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":32,"readmeContent":33,"aiSummary":34,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":35,"discoverSource":36},1173,"shannon","KeygraphHQ\u002Fshannon","KeygraphHQ","Shannon Lite is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production.","https:\u002F\u002Fkeygraph.io\u002F",null,"TypeScript",44487,5155,216,2,0,33,253,2742,166,45,"GNU Affero General Public License v3.0",false,"main",true,[27,28,29,30,31],"penetration-testing","pentesting","security-audit","security-automation","security-tools","2026-06-12 02:00:24",">[!NOTE]\n> **[📢 Sunsetting Router Mode (claude-code-router)`. →](https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fshannon\u002Fdiscussions\u002F301)**\n\n\u003Cdiv align=\"center\">\n\n\u003Cimg src=\".\u002Fassets\u002Fgithub-banner.png\" alt=\"Shannon — AI Pentester for Web Applications and APIs\" width=\"100%\">\n\n# Shannon — AI Pentester by Keygraph\n\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F15604\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Ftrendshift.io\u002Fapi\u002Fbadge\u002Frepositories\u002F15604\" alt=\"KeygraphHQ%2Fshannon | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\nShannon is an autonomous, white-box AI pentester for web applications and APIs. \u003Cbr \u002F>\nIt analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production.\n\n---\n\n\u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002F9ZqQPuhJB7\">\u003Cimg src=\".\u002Fassets\u002Fdiscord.png\" height=\"40\" alt=\"Join Discord\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fkeygraph.io\u002F\">\u003Cimg src=\".\u002Fassets\u002FKeygraph_Button.png\" height=\"40\" alt=\"Visit Keygraph.io\">\u003C\u002Fa>\n\n---\n\u003C\u002Fdiv>\n\n## What is Shannon?\n\nShannon is an AI pentester developed by [Keygraph](https:\u002F\u002Fkeygraph.io). It performs white-box security testing of web applications and their underlying APIs by combining source code analysis with live exploitation.\n\nShannon analyzes your web application's source code to identify potential attack vectors, then uses browser automation and command-line tools to execute real exploits (injection attacks, authentication bypass, SSRF, XSS) against the running application and its APIs. Only vulnerabilities with a working proof-of-concept are included in the final report.\n\n**Why Shannon Exists**\n\nThanks to tools like Claude Code and Cursor, your team ships code non-stop. But your penetration test? That happens once a year. This creates a *massive* security gap. For the other 364 days, you could be unknowingly shipping vulnerabilities to production.\n\nShannon closes that gap by providing on-demand, automated penetration testing that can run against every build or release.\n\n## Shannon in Action\n\nShannon identified 20+ vulnerabilities in OWASP Juice Shop, including authentication bypass and database exfiltration. [Full report →](sample-reports\u002Fshannon-report-juice-shop.md)\n\n![Demo](assets\u002Fshannon-action.gif)\n\n## Features\n\n- **Fully Autonomous Operation**: A single command launches the full pentest. Shannon handles 2FA\u002FTOTP logins (including SSO), browser navigation, exploitation, and report generation without manual intervention.\n- **Reproducible Proof-of-Concept Exploits**: The final report contains only proven, exploitable findings with copy-and-paste PoCs. Vulnerabilities that cannot be exploited are not reported.\n- **OWASP Vulnerability Coverage**: Identifies and validates Injection, XSS, SSRF, and Broken Authentication\u002FAuthorization, with additional categories in development.\n- **Code-Aware Dynamic Testing**: Analyzes source code to guide attack strategy, then validates findings with live browser and CLI-based exploits against the running application.\n- **Parallel Processing**: Vulnerability analysis and exploitation phases run concurrently across all attack categories.\n\n## Product Line\n\nShannon is developed by [Keygraph](https:\u002F\u002Fkeygraph.io) and available in two editions:\n\n| Edition | License | Best For |\n|---------|---------|----------|\n| **Shannon Lite** | AGPL-3.0 | Local testing of your own applications. |\n| **Shannon Pro** | Commercial | Organizations needing a single AppSec platform (SAST, SCA, secrets, business logic testing, autonomous pentesting) with CI\u002FCD integration and self-hosted deployment. |\n\n> **This repository contains Shannon Lite,** the core autonomous AI pentesting framework. **Shannon Pro** is Keygraph's all-in-one AppSec platform, combining SAST, SCA, secrets scanning, business logic security testing, and autonomous AI pentesting in a single correlated workflow. Every finding is validated with a working proof-of-concept exploit.\n\n> [!IMPORTANT]\n> **White-box only.** Shannon Lite is designed for **white-box (source-available)** application security testing.  \n> It expects access to your application's source code and repository layout.\n\n### Shannon Pro: Architecture Overview\n\nShannon Pro is an all-in-one application security platform that replaces the need to stitch together separate SAST, SCA, secrets scanning, and pentesting tools. It operates as a two-stage pipeline: agentic static analysis of the codebase, followed by autonomous AI penetration testing. Findings from both stages are cross-referenced and correlated, so every reported vulnerability has a working proof-of-concept exploit and a precise source code location.\n\n**Stage 1: Agentic Static Analysis**\n\nShannon Pro transforms the codebase into a Code Property Graph (CPG) combining the AST, control flow graph, and program dependence graph. It then runs five analysis capabilities:\n\n- **Data Flow Analysis (SAST)**: Identifies sources (user input, API requests) and sinks (SQL queries, command execution), then traces paths between them. At each node, an LLM evaluates whether the specific sanitization applied is sufficient for the specific vulnerability in context, rather than relying on a hard-coded allowlist of safe functions.\n- **Point Issue Detection (SAST)**: LLM-based detection of single-location vulnerabilities: weak cryptography, hardcoded credentials, insecure configuration, missing security headers, weak RNG, disabled certificate validation, and overly permissive CORS.\n- **Business Logic Security Testing (SAST)**: LLM agents analyze the codebase to discover application-specific invariants (e.g., \"document access must verify organizational ownership\"), generate targeted fuzzers to violate those invariants, and synthesize full PoC exploits. This catches authorization failures and domain-specific logic errors that pattern-based scanners cannot detect.\n- **SCA with Reachability Analysis**: Goes beyond flagging CVEs by tracing whether the vulnerable function is actually reachable from application entry points via the CPG. Unreachable vulnerabilities are deprioritized.\n- **Secrets Detection**: Combines regex pattern matching with LLM-based detection (for dynamically constructed credentials, custom formats, obfuscated tokens) and performs liveness validation against the corresponding service using read-only API calls.\n\n**Stage 2: Autonomous Dynamic Penetration Testing**\n\nThe same multi-agent pentest pipeline as Shannon Lite (reconnaissance, parallel vulnerability analysis, parallel exploitation, reporting), enhanced with static findings injected into the exploitation queue. Static findings are mapped to Shannon's five attack domains (Injection, XSS, SSRF, Auth, Authz), and exploit agents attempt real proof-of-concept attacks against the running application for each finding.\n\n**Static-Dynamic Correlation**\n\nThis is the core differentiator. A data flow vulnerability identified in static analysis (e.g., unsanitized input reaching a SQL query) is not reported as a theoretical risk. It is fed to the corresponding exploit agent, which attempts to exploit it against the live application. Confirmed exploits are traced back to the exact source code location, giving developers both proof of exploitability and the line of code to fix.\n\n**Deployment Model**\n\nShannon Pro supports a self-hosted runner model (similar to GitHub Actions self-hosted runners). The data plane, which handles code access and all LLM API calls, runs entirely within the customer's infrastructure using the customer's own API keys. Source code never leaves the customer's network. The Keygraph control plane handles job orchestration, scan scheduling, and the reporting UI, receiving only aggregate findings.\n\n| Capability | Shannon Lite | Shannon Pro (All-in-One AppSec) |\n| --- | --- | --- |\n| **Licensing** | AGPL-3.0 | Commercial |\n| **Static Analysis** | Code review prompting | Full agentic SAST, SCA, secrets, business logic testing |\n| **Dynamic Testing** | Autonomous AI pentesting | Autonomous AI pentesting with static-dynamic correlation |\n| **Analysis Engine** | Code review prompting | CPG-based data flow with LLM reasoning at every node |\n| **Business Logic** | None | Automated invariant discovery, fuzzer generation, exploit synthesis |\n| **CI\u002FCD Integration** | Manual \u002F CLI | Native CI\u002FCD, GitHub PR scanning |\n| **Deployment** | CLI | Managed cloud or self-hosted runner |\n| **Boundary Analysis** | None | Automatic service boundary detection with team routing |\n\n[Full technical details →](.\u002FSHANNON-PRO.md)\n\n## Table of Contents\n\n- [What is Shannon?](#what-is-shannon)\n- [Shannon in Action](#shannon-in-action)\n- [Features](#features)\n- [Product Line](#product-line)\n- [Setup & Usage Instructions](#setup--usage-instructions)\n  - [Prerequisites](#prerequisites)\n  - [Quick Start (Recommended: npx)](#quick-start-recommended-npx)\n  - [Clone and Build](#clone-and-build)\n  - [Prepare Your Repository](#prepare-your-repository)\n  - [Common Commands](#common-commands)\n  - [Workspaces and Resuming](#workspaces-and-resuming)\n  - [Credentials and Configuration](#credentials-and-configuration)\n  - [AWS Bedrock](#aws-bedrock)\n  - [Google Vertex AI](#google-vertex-ai)\n  - [Custom Base URL](#custom-base-url)\n  - [Platform-Specific Instructions](#platform-specific-instructions)\n  - [Output and Results](#output-and-results)\n- [Sample Reports](#sample-reports)\n- [Benchmark](#benchmark)\n- [Architecture](#architecture)\n- [Coverage and Roadmap](#coverage-and-roadmap)\n- [Disclaimers](#disclaimers)\n- [License](#license)\n- [Community & Support](#community--support)\n- [Get in Touch](#get-in-touch)\n\n---\n\n## Setup & Usage Instructions\n\n### Prerequisites\n\n- **Docker** - Container runtime ([Install Docker](https:\u002F\u002Fdocs.docker.com\u002Fget-docker\u002F))\n- **Node.js 18+** - Required for `npx` usage ([Install Node.js](https:\u002F\u002Fnodejs.org\u002F))\n- **pnpm** - Required for Clone and Build mode ([Install pnpm](https:\u002F\u002Fpnpm.io\u002Finstallation))\n- **AI Provider Credentials** (choose one):\n  - **Anthropic API key** (recommended) - Get from [Anthropic Console](https:\u002F\u002Fconsole.anthropic.com)\n  - **Claude Code OAuth token**\n  - **AWS Bedrock** - Route through Amazon Bedrock with AWS credentials (see [AWS Bedrock](#aws-bedrock))\n  - **Google Vertex AI** - Route through Google Cloud Vertex AI (see [Google Vertex AI](#google-vertex-ai))\n\n> [!NOTE]\n> Docker is still required to use the `npx` workflow. Under the hood, the CLI pulls and runs a prebuilt Shannon worker image from Docker Hub, which is approximately 1 GB and contains Shannon plus all required dependencies. Shannon mounts the target repository as read-only inside the worker container to protect against accidental modifications during analysis. Run Shannon via `npx @keygraph\u002Fshannon` for the latest released version, or pull the latest `main` if building from source.\n\n### Quick Start (Recommended: npx)\n\n> [!WARNING]\n> **Please read the [Disclaimers](#disclaimers) before running Shannon.** Shannon is **not** a passive scanner — it actively executes exploits against the target. You must have **explicit, written authorization** from the system owner.\n\n```bash\n# 1. Configure credentials (interactive wizard — one-time setup)\nnpx @keygraph\u002Fshannon setup\n\n# Or export env vars directly\nexport ANTHROPIC_API_KEY=your-api-key\n\n# 2. Run a pentest\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fyour-app.com -r \u002Fpath\u002Fto\u002Fyour-repo\n```\n\nShannon will pull the worker image from Docker Hub, start the infrastructure, and launch an ephemeral worker container for the scan.\n\n### Clone and Build\n\nUse this if you want to run Shannon from a local clone, modify Shannon itself, or keep the worker image built locally.\n\n```bash\n# 1. Clone Shannon\ngit clone https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fshannon.git\ncd shannon\n\n# 2. Configure credentials (choose one method)\n\n# Option A: Create a .env file\ncat > .env \u003C\u003C 'EOF'\nANTHROPIC_API_KEY=your-api-key\nCLAUDE_CODE_MAX_OUTPUT_TOKENS=64000\nEOF\n\n# Option B: Export environment variables\nexport ANTHROPIC_API_KEY=\"your-api-key\"              # or CLAUDE_CODE_OAUTH_TOKEN\nexport CLAUDE_CODE_MAX_OUTPUT_TOKENS=64000           # recommended\n\n# 3. Install dependencies and build\npnpm install\npnpm build\n\n# 4. Run a pentest\n.\u002Fshannon start -u https:\u002F\u002Fyour-app.com -r \u002Fpath\u002Fto\u002Fyour-repo\n```\n\nShannon will build the worker image locally, start the infrastructure, and launch an ephemeral worker container for the scan.\n\n### Prepare Your Repository\n\nShannon can scan any repository on your machine. Pass an absolute or relative path with `-r`.\n\nExamples:\n\n```bash\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r .\u002Frelative\u002Fpath\n```\n\n\u003C\u002Fdetails>\n\n### Common Commands\n\n#### Monitoring Progress\n\n```bash\nnpx @keygraph\u002Fshannon logs \u003Cworkspace>\nnpx @keygraph\u002Fshannon status\n```\n\nOpen the Temporal Web UI for detailed monitoring:\n\n```bash\nopen http:\u002F\u002Flocalhost:8233\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon logs \u003Cworkspace>\n.\u002Fshannon status\n```\n\n\u003C\u002Fdetails>\n\n#### Stopping Shannon\n\n```bash\nnpx @keygraph\u002Fshannon stop\nnpx @keygraph\u002Fshannon stop --clean\nnpx @keygraph\u002Fshannon uninstall\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon stop\n.\u002Fshannon stop --clean\n```\n\n\u003C\u002Fdetails>\n\n#### Usage Examples\n\n```bash\n# Basic pentest\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo\n\n# With a configuration file\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -c \u002Fpath\u002Fto\u002Fmy-config.yaml\n\n# Custom output directory\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -o .\u002Fmy-reports\n\n# Named workspace\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w q1-audit\n\n# List all workspaces\nnpx @keygraph\u002Fshannon workspaces\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n# Basic pentest\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo\n\n# With a configuration file\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -c \u002Fpath\u002Fto\u002Fmy-config.yaml\n\n# Custom output directory\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -o .\u002Fmy-reports\n\n# Named workspace\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w q1-audit\n\n# List all workspaces\n.\u002Fshannon workspaces\n\n# Rebuild worker image\n.\u002Fshannon build --no-cache\n```\n\n\u003C\u002Fdetails>\n\n### Workspaces and Resuming\n\nShannon supports **workspaces** that allow you to resume interrupted or failed runs without re-running completed agents.\n\n**How it works:**\n\n- Every run creates a workspace (auto-named by default, for example `example-com_shannon-1771007534808`)\n- Workspaces are stored in `.\u002Fworkspaces\u002F` (local mode) or `~\u002F.shannon\u002Fworkspaces\u002F` (npx mode)\n- Use `-w \u003Cname>` to give your run a custom name for easier reference\n- To resume any run, pass its workspace name via `-w` — Shannon detects which agents completed successfully and picks up where it left off\n- Each agent's progress is checkpointed via git commits, so resumed runs start from a clean, validated state\n\n```bash\n# Start with a named workspace\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w my-audit\n\n# Resume the same workspace (skips completed agents)\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w my-audit\n\n# Resume an auto-named workspace from a previous run\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w example-com_shannon-1771007534808\n\n# List all workspaces and their status\nnpx @keygraph\u002Fshannon workspaces\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w my-audit\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w my-audit\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -w example-com_shannon-1771007534808\n.\u002Fshannon workspaces\n```\n\n\u003C\u002Fdetails>\n\n> [!NOTE]\n> The `URL` must match the original workspace URL when resuming. Shannon will reject mismatched URLs to prevent cross-target contamination.\n\n### Credentials and Configuration\n\n#### Credential Precedence\n\n**Local mode** resolves credentials from:\n\n1. **Environment variables** - `export ANTHROPIC_API_KEY=...`\n2. **`.env` file** - `.\u002F.env`\n\n**npx mode** uses TOML instead of `.env`:\n\n1. **Environment variables** - `export ANTHROPIC_API_KEY=...`\n2. **`~\u002F.shannon\u002Fconfig.toml`** - created by `npx @keygraph\u002Fshannon setup`\n\nEnvironment variables always win, so you can override saved config for a single session without editing files.\n\n#### Configuration (Optional)\n\nWhile you can run without a config file, creating one enables authenticated testing and customized analysis. Pass any configuration file path with `-c`.\n\n##### Create Configuration File\n\nCopy and modify the example configuration:\n\n```bash\ncp configs\u002Fexample-config.yaml .\u002Fmy-app-config.yaml\n```\n\n##### Basic Configuration Structure\n\n```yaml\n# Describe your target environment (optional, max 500 chars)\ndescription: \"Next.js e-commerce app on PostgreSQL. Local dev environment — .env files contain local-only credentials, not deployed to production.\"\n\n# Limit which vulnerability classes run end-to-end (optional, default: all five)\n# vuln_classes: [injection, xss, auth, authz, ssrf]\n\n# Skip the exploitation phase (optional, default: \"true\")\n# exploit: \"false\"\n\n# Free-form rules of engagement (optional)\n# rules_of_engagement: |\n#   - No password brute-force; cap login attempts at 5 per account.\n#   - Throttle to under 5 requests per second per endpoint; back off 60s on any 429.\n#   - Use placeholders like [order_id] in deliverables — no real data values.\n\nauthentication:\n  login_type: form\n  login_url: \"https:\u002F\u002Fyour-app.com\u002Flogin\"\n  credentials:\n    username: \"test@example.com\"\n    password: \"yourpassword\"\n    totp_secret: \"LB2E2RX7XFHSTGCK\"  # Optional for 2FA\n\n  login_flow:\n    - \"Type $username into the email field\"\n    - \"Type $password into the password field\"\n    - \"Click the 'Sign In' button\"\n\n  success_condition:\n    type: url_contains\n    value: \"\u002Fdashboard\"\n\nrules:\n  # Supported types: url_path, subdomain, domain, method, header, parameter, code_path\n  avoid:\n    - description: \"AI should avoid testing logout functionality\"\n      type: url_path\n      value: \"\u002Flogout\"\n\n    # code_path values are repo-relative file paths or globs (e.g. \"src\u002Fauth.ts\", \"src\u002Fvendor\u002F**\").\n    # - description: \"Out-of-scope vendored libraries\"\n    #   type: code_path\n    #   value: \"src\u002Fvendor\u002F**\"\n\n  focus:\n    - description: \"AI should emphasize testing API endpoints\"\n      type: url_path\n      value: \"\u002Fapi\"\n\n# Filters applied by the report agent when assembling the final report (optional).\n# report:\n#   min_severity: low                   # drop findings below this severity (low | medium | high | critical)\n#   min_confidence: low                 # drop findings below this confidence (low | medium | high)\n#   guidance: |\n#     Drop findings about missing security headers and rate-limit gaps.\n```\n\nRun with:\n\n```bash\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -c .\u002Fmy-app-config.yaml\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon start -u https:\u002F\u002Fexample.com -r \u002Fpath\u002Fto\u002Frepo -c .\u002Fmy-app-config.yaml\n```\n\n\u003C\u002Fdetails>\n\n#### TOTP Setup for 2FA\n\nIf your application uses two-factor authentication, simply add the TOTP secret to your config file. The AI will automatically generate the required codes during testing.\n\n#### Adaptive Thinking (Opus 4.6\u002F4.7)\n\nClaude decides when and how deeply to reason on Opus 4.6 and 4.7. Enabled by default whenever a tier resolves to one of these models.\n\n- **npx mode** — `npx @keygraph\u002Fshannon setup` prompts you during the wizard.\n- **Local mode** — set `CLAUDE_ADAPTIVE_THINKING=false` in `.env` (or as an exported env var) to disable.\n\n#### Subscription Plan Rate Limits\n\nAnthropic subscription plans reset usage on a **rolling 5-hour window**. The default retry strategy (30-min max backoff) will exhaust retries before the window resets. Add this to your config:\n\n```yaml\npipeline:\n  retry_preset: subscription          # Extends max backoff to 6h, 100 retries\n  max_concurrent_pipelines: 2         # Run 2 of 5 pipelines at a time (reduces burst API usage)\n```\n\n`max_concurrent_pipelines` controls how many vulnerability pipelines run simultaneously (1-5, default: 5). Lower values reduce the chance of hitting rate limits but increase wall-clock time.\n\n### AWS Bedrock\n\nShannon also supports [Amazon Bedrock](https:\u002F\u002Faws.amazon.com\u002Fbedrock\u002F) instead of using an Anthropic API key.\n\n#### Quick Setup\n\nRun `npx @keygraph\u002Fshannon setup` and select **AWS Bedrock**. The wizard will prompt for your region, bearer token, and model IDs.\n\nOr export env vars directly:\n\n```bash\nexport CLAUDE_CODE_USE_BEDROCK=1\nexport AWS_REGION=us-east-1\nexport AWS_BEARER_TOKEN_BEDROCK=your-bearer-token\nexport ANTHROPIC_SMALL_MODEL=us.anthropic.claude-haiku-4-5-20251001-v1:0\nexport ANTHROPIC_MEDIUM_MODEL=us.anthropic.claude-sonnet-4-6\nexport ANTHROPIC_LARGE_MODEL=us.anthropic.claude-opus-4-7\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build: add to .env instead\u003C\u002Fsummary>\n\n```bash\nCLAUDE_CODE_USE_BEDROCK=1\nAWS_REGION=us-east-1\nAWS_BEARER_TOKEN_BEDROCK=your-bearer-token\nANTHROPIC_SMALL_MODEL=us.anthropic.claude-haiku-4-5-20251001-v1:0\nANTHROPIC_MEDIUM_MODEL=us.anthropic.claude-sonnet-4-6\nANTHROPIC_LARGE_MODEL=us.anthropic.claude-opus-4-7\n```\n\n\u003C\u002Fdetails>\n\nShannon uses three model tiers: **small** (`claude-haiku-4-5-20251001`) for summarization, **medium** (`claude-sonnet-4-6`) for security analysis, and **large** (`claude-opus-4-7`) for deep reasoning. Set `ANTHROPIC_SMALL_MODEL`, `ANTHROPIC_MEDIUM_MODEL`, and `ANTHROPIC_LARGE_MODEL` to the Bedrock model IDs for your region.\n\n### Google Vertex AI\n\nShannon also supports [Google Vertex AI](https:\u002F\u002Fcloud.google.com\u002Fvertex-ai) instead of using an Anthropic API key.\n\nCreate a service account with the `roles\u002Faiplatform.user` role in the [GCP Console](https:\u002F\u002Fconsole.cloud.google.com\u002Fiam-admin\u002Fserviceaccounts), then download a JSON key file.\n\n#### Quick Setup\n\nRun `npx @keygraph\u002Fshannon setup` and select **Google Vertex AI**. The wizard will prompt for your region, project ID, service account key file path, and model IDs. The key file is securely copied to `~\u002F.shannon\u002Fgoogle-sa-key.json`.\n\nOr export env vars directly:\n\n```bash\nexport CLAUDE_CODE_USE_VERTEX=1\nexport CLOUD_ML_REGION=us-east5\nexport ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id\nexport GOOGLE_APPLICATION_CREDENTIALS=\u002Fpath\u002Fto\u002Fyour-sa-key.json\nexport ANTHROPIC_SMALL_MODEL=claude-haiku-4-5@20251001\nexport ANTHROPIC_MEDIUM_MODEL=claude-sonnet-4-6\nexport ANTHROPIC_LARGE_MODEL=claude-opus-4-7\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build: add to .env instead\u003C\u002Fsummary>\n\n```bash\nCLAUDE_CODE_USE_VERTEX=1\nCLOUD_ML_REGION=us-east5\nANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id\nGOOGLE_APPLICATION_CREDENTIALS=.\u002Fcredentials\u002Fgoogle-sa-key.json\nANTHROPIC_SMALL_MODEL=claude-haiku-4-5@20251001\nANTHROPIC_MEDIUM_MODEL=claude-sonnet-4-6\nANTHROPIC_LARGE_MODEL=claude-opus-4-7\n```\n\n\u003C\u002Fdetails>\n\nSet `CLOUD_ML_REGION=global` for global endpoints, or a specific region like `us-east5`. Some models may not be available on global endpoints — see the [Vertex AI Model Garden](https:\u002F\u002Fconsole.cloud.google.com\u002Fvertex-ai\u002Fmodel-garden) for region availability.\n\n### Custom Base URL\n\nShannon supports pointing the SDK at any Anthropic-compatible endpoint via `ANTHROPIC_BASE_URL`. For users who need proxy-based routing, the supported path is to use an LLM proxy such as [LiteLLM](https:\u002F\u002Fgithub.com\u002FBerriAI\u002Flitellm) configured to expose an Anthropic-compatible endpoint.\n\n> [!IMPORTANT]\n> **Only Claude models are officially supported.** Shannon's evaluations, internal testing, and agent harness are all optimized for Claude. Smaller or alternative models — including non-Claude models routed through a proxy — may not reliably follow Shannon's instructions or tool-use constraints, and are not officially supported. Use them at your own risk; results may be incomplete, inaccurate, or unstable.\n>\n> The previously experimental `claude-code-router` integration is being removed in an upcoming release. If you currently rely on it, migrate to an Anthropic-compatible proxy such as LiteLLM before upgrading.\n\nRun `npx @keygraph\u002Fshannon setup` and select **Custom Base URL**. The wizard will prompt for your endpoint URL, auth token, and optionally let you override the default model tiers.\n\nOr export env vars directly:\n\n```bash\nexport ANTHROPIC_BASE_URL=https:\u002F\u002Fyour-proxy.example.com\nexport ANTHROPIC_AUTH_TOKEN=your-auth-token\n\n# Optionally override model tiers (defaults are used if not set)\nexport ANTHROPIC_SMALL_MODEL=claude-haiku-4-5-20251001\nexport ANTHROPIC_MEDIUM_MODEL=claude-sonnet-4-6\nexport ANTHROPIC_LARGE_MODEL=claude-opus-4-7\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build: add to .env instead\u003C\u002Fsummary>\n\n```bash\nANTHROPIC_BASE_URL=https:\u002F\u002Fyour-proxy.example.com\nANTHROPIC_AUTH_TOKEN=your-auth-token\nANTHROPIC_SMALL_MODEL=claude-haiku-4-5-20251001\nANTHROPIC_MEDIUM_MODEL=claude-sonnet-4-6\nANTHROPIC_LARGE_MODEL=claude-opus-4-7\n```\n\n\u003C\u002Fdetails>\n\n### Platform-Specific Instructions\n\n**For Windows:**\n\nShannon on Windows is only supported via **WSL2**. Native Windows (including Git Bash) is not supported.\n\n**Step 1: Ensure WSL 2**\n\n```powershell\nwsl --install\nwsl --set-default-version 2\n\n# Check installed distros\nwsl --list --verbose\n\n# If you don't have a distro, install one (Ubuntu 24.04 recommended)\nwsl --list --online\nwsl --install Ubuntu-24.04\n\n# If your distro shows VERSION 1, convert it to WSL 2:\nwsl --set-version \u003Cdistro-name> 2\n```\n\nSee [WSL basic commands](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Fbasic-commands) for reference.\n\n**Step 2: Install Docker Desktop on Windows** and enable **WSL2 backend** under *Settings > General > Use the WSL 2 based engine*.\n\n**Step 3: Run Shannon inside WSL** using either flow.\n\n**npx inside WSL:**\n\n```bash\nnpx @keygraph\u002Fshannon setup\nnpx @keygraph\u002Fshannon start -u https:\u002F\u002Fyour-app.com -r \u002Fpath\u002Fto\u002Fyour-repo\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fshannon.git\ncd shannon\ncp .env.example .env  # Edit with your API key\n.\u002Fshannon start -u https:\u002F\u002Fyour-app.com -r \u002Fpath\u002Fto\u002Fyour-repo\n```\n\n\u003C\u002Fdetails>\n\nTo access the Temporal Web UI, run `ip addr` inside WSL to find your WSL IP address, then navigate to `http:\u002F\u002F\u003Cwsl-ip>:8233` in your Windows browser.\n\nWindows Defender may flag exploit code in reports as false positives; see [Antivirus False Positives](#6-windows-antivirus-false-positives) below.\n\n**For Linux (Native Docker):**\n\nYou may need to run commands with `sudo` depending on your Docker setup. If you encounter permission issues with output files, ensure your user has access to the Docker socket.\n\n**For macOS:**\n\nWorks out of the box with Docker Desktop installed.\n\n**Testing Local Applications:**\n\nDocker containers cannot reach `localhost` on your host machine. Use `host.docker.internal` in place of `localhost`:\n\n```bash\nnpx @keygraph\u002Fshannon start -u http:\u002F\u002Fhost.docker.internal:3000 -r \u002Fpath\u002Fto\u002Frepo\n```\n\n\u003Cdetails>\n\u003Csummary>Clone and Build command equivalents\u003C\u002Fsummary>\n\n```bash\n.\u002Fshannon start -u http:\u002F\u002Fhost.docker.internal:3000 -r \u002Fpath\u002Fto\u002Frepo\n```\n\n\u003C\u002Fdetails>\n\n### Output and Results\n\nAll results are saved to the workspaces directory: `.\u002Fworkspaces\u002F` (local mode) or `~\u002F.shannon\u002Fworkspaces\u002F` (npx mode). Use `-o \u003Cpath>` to copy deliverables to a custom output directory after the run completes.\n\nOutput structure:\n\n```text\nworkspaces\u002F{hostname}_{sessionId}\u002F\n├── session.json          # Metrics and session data\n├── workflow.log          # Human-readable workflow log\n├── agents\u002F               # Per-agent execution logs\n├── prompts\u002F              # Prompt snapshots for reproducibility\n└── deliverables\u002F\n    └── comprehensive_security_assessment_report.md   # Final comprehensive security report\n```\n\n---\n\n## Sample Reports\n\nSample penetration test reports from industry-standard vulnerable applications:\n\n#### **OWASP Juice Shop** • [GitHub](https:\u002F\u002Fgithub.com\u002Fjuice-shop\u002Fjuice-shop)\n\n*A notoriously insecure web application maintained by OWASP, designed to test a tool's ability to uncover a wide range of modern vulnerabilities.*\n\n**Results**: Identified over 20 vulnerabilities across targeted OWASP categories in a single automated run.\n\n**Notable findings**:\n\n- Authentication bypass and full user database exfiltration via SQL injection\n- Privilege escalation to administrator through registration workflow bypass\n- IDOR vulnerabilities enabling access to other users' data and shopping carts\n- SSRF enabling internal network reconnaissance\n\n[View Complete Report →](sample-reports\u002Fshannon-report-juice-shop.md)\n\n---\n\n#### **c{api}tal API** • [GitHub](https:\u002F\u002Fgithub.com\u002FCheckmarx\u002Fcapital)\n\n*An intentionally vulnerable API from Checkmarx, designed to test a tool's ability to uncover the OWASP API Security Top 10.*\n\n**Results**: Identified approximately 15 critical and high-severity vulnerabilities.\n\n**Notable findings**:\n\n- Root-level command injection via denylist bypass in a hidden debug endpoint\n- Authentication bypass through a legacy, unpatched v1 API endpoint\n- Privilege escalation via Mass Assignment in the user profile update function\n- Zero false positives for XSS (correctly confirmed robust XSS defenses)\n\n[View Complete Report →](sample-reports\u002Fshannon-report-capital-api.md)\n\n---\n\n#### **OWASP crAPI** • [GitHub](https:\u002F\u002Fgithub.com\u002FOWASP\u002FcrAPI)\n\n*A modern, intentionally vulnerable API from OWASP, designed to benchmark a tool's effectiveness against the OWASP API Security Top 10.*\n\n**Results**: Identified over 15 critical and high-severity vulnerabilities.\n\n**Notable findings**:\n\n- Authentication bypass via multiple JWT attacks (Algorithm Confusion, alg:none, weak key injection)\n- Full PostgreSQL database compromise via injection, exfiltrating user credentials\n- SSRF attack forwarding internal authentication tokens to an external service\n- Zero false positives for XSS (correctly identified robust XSS defenses)\n\n[View Complete Report →](sample-reports\u002Fshannon-report-crapi.md)\n\n---\n\n## Benchmark\n\nShannon Lite scored **96.15% (100\u002F104 exploits)** on a hint-free, source-aware variant of the XBOW security benchmark.\n\n**[Full results with detailed agent logs and per-challenge pentest reports →](https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fxbow-validation-benchmarks\u002Fblob\u002Fmain\u002Fxben-benchmark-results\u002F)**\n\n---\n\n## Architecture\n\nShannon uses a multi-agent architecture that combines white-box source code analysis with dynamic exploitation across five phases:\n\n```\n        ┌──────────────────────┐\n        │   Pre-Reconnaissance │\n        │   (source code scan) │\n        └──────────┬───────────┘\n                   │\n                   ▼\n        ┌──────────────────────┐\n        │   Reconnaissance     │\n        │  (attack surface     │\n        │   mapping)           │\n        └──────────┬───────────┘\n                   │\n                   ▼\n        ┌──────────┴───────────┐\n        │          │           │\n        ▼          ▼           ▼\n  ┌───────────┐ ┌───────────┐ ┌───────────┐\n  │ Vuln      │ │ Vuln      │ │   ...     │\n  │(Injection)│ │  (XSS)    │ │           │\n  └─────┬─────┘ └─────┬─────┘ └─────┬─────┘\n        │              │             │\n        ▼              ▼             ▼\n  ┌───────────┐ ┌───────────┐ ┌───────────┐\n  │ Exploit   │ │ Exploit   │ │   ...     │\n  │(Injection)│ │  (XSS)    │ │           │\n  └─────┬─────┘ └─────┬─────┘ └─────┬─────┘\n        │              │             │\n        └──────┬───────┴─────────────┘\n               │\n               ▼\n        ┌──────────────────────┐\n        │      Reporting       │\n        └──────────────────────┘\n```\n\n### Architectural Overview\n\nShannon uses Anthropic's Claude Agent SDK as its reasoning engine within a multi-agent architecture. The system combines white-box source code analysis with black-box dynamic exploitation, managed by an orchestrator across five phases. The architecture is designed for minimal false positives through a \"no exploit, no report\" policy.\n\nEach scan runs in its own ephemeral Docker container (`docker run --rm`) with a per-invocation Temporal task queue, enabling concurrent scans with different target repositories.\n\n---\n\n#### **Phase 1: Pre-Reconnaissance**\n\nPerforms source code analysis to identify the application framework, entry points, and potential attack surface from the codebase. Builds the foundational architectural intelligence that all subsequent agents depend on.\n\n#### **Phase 2: Reconnaissance**\n\nBuilds a comprehensive attack surface map from the pre-recon findings. Shannon performs live application exploration via browser automation to correlate code-level insights with real-world behavior, producing a detailed map of all entry points, API endpoints, and authentication mechanisms.\n\n#### **Phase 3: Vulnerability Analysis**\n\nTo maximize efficiency, this phase operates in parallel with 5 concurrent agents. Using the reconnaissance data, specialized agents for each OWASP category (injection, XSS, auth, authz, SSRF) hunt for potential flaws in parallel. For vulnerabilities like Injection and SSRF, agents perform a structured data flow analysis, tracing user input to dangerous sinks. This phase produces a key deliverable: a list of **hypothesized exploitable paths** that are passed on for validation.\n\n#### **Phase 4: Exploitation**\n\nContinuing the parallel workflow to maintain speed, this phase is dedicated entirely to turning hypotheses into proof. Dedicated exploit agents receive the hypothesized paths and attempt to execute real-world attacks using browser automation, command-line tools, and custom scripts. This phase enforces a strict **\"No Exploit, No Report\"** policy: if a hypothesis cannot be successfully exploited to demonstrate impact, it is discarded as a false positive.\n\n#### **Phase 5: Reporting**\n\nThe final phase compiles all validated findings into a professional, actionable report. An agent consolidates the reconnaissance data and the successful exploit evidence, cleaning up any noise or hallucinated artifacts. Only verified vulnerabilities are included, complete with **reproducible, copy-and-paste Proof-of-Concepts**, delivering a final pentest-grade report focused exclusively on proven risks.\n\n\n## Coverage and Roadmap\n\nFor detailed information about Shannon's security testing coverage and development roadmap, see our [Coverage and Roadmap](.\u002FCOVERAGE.md) documentation.\n\n## Disclaimers\n\n### Important Usage Guidelines & Disclaimers\n\nPlease review the following guidelines carefully before using Shannon (Lite). As a user, you are responsible for your actions and assume all liability.\n\n#### **1. Potential for Mutative Effects & Environment Selection**\n\nThis is not a passive scanner. The exploitation agents are designed to **actively execute attacks** to confirm vulnerabilities. This process can have mutative effects on the target application and its data.\n\n> [!WARNING]\n> **DO NOT run Shannon on production environments.**\n>\n> - It is intended exclusively for use on sandboxed, staging, or local development environments where data integrity is not a concern.\n> - Potential mutative effects include, but are not limited to: creating new users, modifying or deleting data, compromising test accounts, and triggering unintended side effects from injection attacks.\n> - **For maximum security and isolation, run Shannon inside a virtual machine (VM).** This confines any side effects from exploitation — including unexpected outbound traffic, file writes from agent tooling, or interactions with local services — to a disposable environment.\n\n#### **2. Legal & Ethical Use**\n\nShannon is designed for legitimate security auditing purposes only.\n\n> [!CAUTION]\n> **You must have explicit, written authorization** from the owner of the target system before running Shannon.\n>\n> Unauthorized scanning and exploitation of systems you do not own is illegal and can be prosecuted under laws such as the Computer Fraud and Abuse Act (CFAA). Keygraph is not responsible for any misuse of Shannon.\n\n#### **3. LLM & Automation Caveats**\n\n- **Verification is Required**: While significant engineering has gone into our \"proof-by-exploitation\" methodology to eliminate false positives, the underlying LLMs can still generate hallucinated or weakly-supported content in the final report. **Human oversight is essential** to validate the legitimacy and severity of all reported findings.\n- **Model Support**: Shannon is officially supported only with **Claude models**. Our evaluations, internal testing, and agent harness are all optimized for Claude. Smaller or alternative models — including non-Claude models routed through a proxy — may not reliably follow Shannon's instructions or tool-use constraints, and are not officially supported.\n- **Comprehensiveness**: The analysis in Shannon Lite may not be exhaustive due to the inherent limitations of LLM context windows. For a more comprehensive, graph-based analysis of your entire codebase, **Shannon Pro** leverages its advanced data flow analysis engine to ensure deeper and more thorough coverage.\n\n#### **4. Scope of Analysis**\n\n- **Targeted Vulnerabilities**: The current version of Shannon Lite specifically targets the following classes of *exploitable* vulnerabilities:\n  - Broken Authentication & Authorization\n  - Injection\n  - Cross-Site Scripting (XSS)\n  - Server-Side Request Forgery (SSRF)\n- **What Shannon Lite Does Not Cover**: This list is not exhaustive of all potential security risks. Shannon Lite's \"proof-by-exploitation\" model means it will not report on issues it cannot actively exploit, such as vulnerable third-party libraries or insecure configurations. These types of deep static-analysis findings are a core focus of the advanced analysis engine in **Shannon Pro**.\n\n#### **5. Cost & Performance**\n\n- **Time**: As of the current version, a full test run typically takes **1 to 1.5 hours** to complete.\n- **Cost**: Running the full test using Anthropic's Claude 4.5 Sonnet model may incur costs of approximately **$50 USD**. Costs vary based on model pricing and application complexity.\n\n#### **6. Windows Antivirus False Positives**\n\nWindows Defender may flag files in `xben-benchmark-results\u002F` or `deliverables\u002F` as malware. These are false positives caused by exploit code in the reports. Add an exclusion for the Shannon directory in Windows Defender, or use Docker\u002FWSL2.\n\n#### **7. Security Considerations**\n\nShannon Lite is designed for scanning repositories and applications you own or have explicit permission to test. Do not point it at untrusted or adversarial codebases. Like any AI-powered tool that reads source code, Shannon Lite is susceptible to prompt injection from content in the scanned repository.\n\n\n## License\n\nShannon Lite is released under the [GNU Affero General Public License v3.0 (AGPL-3.0)](LICENSE).\n\nShannon is open source (AGPL v3). This license allows you to:\n- Use it freely for all internal security testing.\n- Modify the code privately for internal use without sharing your changes.\n\nThe AGPL's sharing requirements primarily apply to organizations offering Shannon as a public or managed service (such as a SaaS platform). In those specific cases, any modifications made to the core software must be open-sourced.\n\n\n## Community & Support\n\n### Community Resources\n\n**1:1 Office Hours** — Thursdays, two time zones\nBook a free 15-min session for hands-on help with bugs, deployments, or config questions.\n→ US\u002FEU: 10:00 AM PT  |  Asia: 2:00 PM IST\n→ [Book a slot](https:\u002F\u002Fcal.com\u002Fgeorge-flores-keygraph\u002Fshannon-community-office-hours)\n\n[Join our Discord](https:\u002F\u002Fdiscord.gg\u002FcmctpMBXwE) to ask questions, share feedback, and connect with other Shannon users.\n\n**Contributing:** At this time, we're not accepting external code contributions (PRs).  \nIssues are welcome for bug reports and feature requests.\n\n- **Report bugs** via [GitHub Issues](https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fshannon\u002Fissues)\n- **Suggest features** in [Discussions](https:\u002F\u002Fgithub.com\u002FKeygraphHQ\u002Fshannon\u002Fdiscussions)\n\n### Stay Connected\n\n- **Twitter**: [@KeygraphHQ](https:\u002F\u002Ftwitter.com\u002FKeygraphHQ)\n- **LinkedIn**: [Keygraph](https:\u002F\u002Flinkedin.com\u002Fcompany\u002Fkeygraph)\n- **Website**: [keygraph.io](https:\u002F\u002Fkeygraph.io)\n\n\n\n## Get in Touch\n\n### Shannon Pro\n\nShannon Pro is Keygraph's all-in-one AppSec platform. For organizations that need unified SAST, SCA, and autonomous pentesting with static-dynamic correlation, CI\u002FCD integration, or self-hosted deployment, see the [Shannon Pro technical overview](.\u002FSHANNON-PRO.md).\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fcal.com\u002Fteam\u002Fkeygraph\u002Fshannon-pro\" target=\"_blank\">\n    \u003Cimg src=\".\u002Fassets\u002FDemo_Button.png\" height=\"40\" alt=\"Shannon Pro Inquiry\">\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\n**Email**: [shannon@keygraph.io](mailto:shannon@keygraph.io)\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cb>Built by \u003Ca href=\"https:\u002F\u002Fkeygraph.io\">Keygraph\u003C\u002Fa>\u003C\u002Fb>\n\u003C\u002Fp>\n","Shannon 是一个针对Web应用和API的自主白盒AI渗透测试工具。它通过分析源代码来识别攻击向量，并执行真实的漏洞利用以在生产环境之前验证漏洞。项目采用TypeScript编写，具备全自动操作、可重现的漏洞利用证明、OWASP漏洞覆盖以及基于代码的动态测试等核心功能。适用于需要频繁进行安全测试的开发团队，尤其是在持续集成\u002F持续部署（CI\u002FCD）流程中，确保每次构建或发布都能自动检查并报告潜在的安全风险。","2026-06-11 02:42:08","top_all"]