affaan-m/agentshield

How affaan-m/agentshield works

AgentShield audits AI-agent configuration, chiefly by discovering local configuration, applying security rules, grading findings, and presenting reports. The CLI also exposes remediation, policy and baseline workflows, evidence bundles, monitoring, and optional LLM-assisted analysis; MiniClaw is a separate sandboxed agent service. Edges reflect the supplied excerpts, SOURCE INDEX references, and README. Several documented capabilities have no sampled implementation details, so their internal connections are omitted.

Main entry points: src/index.ts CLI · scanner/index.ts Discovery and Scan · server.ts MiniClaw Server · pipeline.ts Opus Analysis

Components 25

Configuration Scanning

Discovery and Scan src/scanner/index.ts
Config Discovery src/scanner/discovery.ts
Rule Registry src/rules/index.ts
Security Rules src/rules
Harness Adapters src/harness-adapters/index.ts
Skill Health src/skills/health.ts

Findings and Remediation

Scoring src/reporter/score.ts
Report Formats src/reporter
Auto-fix src/fixer/index.ts

Audit Governance

Policy Evaluation src/policy
Baseline Comparison src/baseline
Evidence Pack src/evidence-pack/index.ts

Advanced Analysis

Supply-chain Analysis src/supply-chain
Opus Analysis src/opus/pipeline.ts
LLM Client src/llm/client.ts
Injection Testing src/injection

Agent Operations

Configuration Watch src/watch
Runtime Monitor src/runtime
MiniClaw Server src/miniclaw/server.ts
Prompt Router src/miniclaw/router.ts
MiniClaw Sandbox src/miniclaw/sandbox.ts
Safe Tools src/miniclaw/tools.ts

Other components

Developer
CLI src/index.ts
Anthropic API

Connections 29

Developer → CLI: runs commands README.md
CLI → Discovery and Scan: starts scan src/index.ts
Discovery and Scan → Config Discovery: discovers files src/scanner/index.ts
Discovery and Scan → Rule Registry: loads rules src/scanner/index.ts
Rule Registry → Security Rules: registers checks src/rules/index.ts
Discovery and Scan → Harness Adapters: detects adapters src/scanner/index.ts
Discovery and Scan → Skill Health: analyzes skills src/scanner/index.ts
CLI → Scoring: scores findings src/index.ts
CLI → Report Formats: renders reports src/index.ts
CLI → Auto-fix: applies fixes src/index.ts
CLI → Policy Evaluation: evaluates policy src/index.ts
CLI → Baseline Comparison: compares baseline src/index.ts
CLI → Evidence Pack: writes bundle src/index.ts
Evidence Pack → Report Formats: renders artifacts src/evidence-pack/index.ts
CLI → Supply-chain Analysis: analyzes supply chain src/index.ts
CLI → Opus Analysis: runs deep analysis src/index.ts
Opus Analysis → Discovery and Scan: uses scan result src/opus/pipeline.ts
Opus Analysis → LLM Client: creates client src/opus/pipeline.ts
LLM Client → Anthropic API: sends requests src/llm/client.ts
CLI → Injection Testing: runs injection tests src/index.ts
Injection Testing → LLM Client: uses client src/injection/index.ts
CLI → Configuration Watch: starts watching src/index.ts
CLI → Runtime Monitor: installs monitor src/index.ts
Developer → MiniClaw Server: sends prompts
MiniClaw Server → MiniClaw Sandbox: creates sandbox src/miniclaw/server.ts
MiniClaw Server → Prompt Router: routes prompts src/miniclaw/server.ts
MiniClaw Server → Safe Tools: builds whitelist src/miniclaw/server.ts
Prompt Router → MiniClaw Sandbox: uses sandbox src/miniclaw/router.ts
Safe Tools → MiniClaw Sandbox: uses sandbox src/miniclaw/tools.ts

Drawn by GitDiagram on November 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·