AgentGuard — Development skill for Claude Code
Guardian Agent and Token Savings for Claude Code.
How to install AgentGuard
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open quilrai/AgentGuard and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What AgentGuard does
Guardian Agent and Token Savings for Claude Code.
Alternatives in Development
- GenericAgent — Self-evolving agent: grows skill tree from 3.3K-line seed, achieving full system control with 6x less token co 6.1k ★
- Claude Token Efficient — One CLAUDE.md file 3.8k ★
- Doctor — Environment diagnostics with interactive fixes — providers, auth, RTK, hooks, token optimization 2.8k ★
README
AgentGuard
Local guardrails and token saving for coding agents.
Website: https://agentguard.quilrai.dev
Downloads:
Windows build: coming soon.

What It Is
AgentGuard is a Tauri desktop app. It runs locally, listens on `localhost:8008`, and connects to agent CLIs/IDEs through hooks.
No proxy. No cloud relay. Hook events come in, policy decisions go out.
Supports:
- Claude Code
- Codex CLI
- Cursor IDE
Save Costs With Claude Code
AgentGuard cuts token waste before it reaches Claude Code.
- Smart file-read cache: unchanged files come back as compact references instead of being resent.
- Context-aware reads: full, diff, or line-range responses are chosen based on what changed.
- Shell output compression: noisy output from builds, tests, package managers, logs,
git,grep,rg,find,ls,curl, and more gets summarized. - Search compression: large
grep/rgresults are grouped by file and trimmed to the useful matches. - Cheaper shell habits:
rg-style search, targeted reads, and summarized file discovery instead of giant recursive dumps. - Diff compression: keeps the actual changes, drops excess context.
- JSON crusher: trims huge arrays, long strings, and deep nesting while preserving structure.
Use the Token Saver tab and apply recommended settings for Claude Code.
Guardian Agent
AgentGuard puts a local policy layer in front of coding agents.
- Sensitive data guard: blocks prompts, tool inputs, file reads, shell commands, and MCP calls that match secrets, credentials, API keys, PII, or custom patterns.
- Dependency guard: blocks known compromised or vulnerable packages before the agent installs them.
- Update advisor: nudges the agent toward newer
Related Skills
Memstack Headroom
Check Headroom proxy status and token savings.
Token Saver Stats
Show token-saver compression statistics and savings
Tldr Stats
Real session token usage + lifetime savings + USD estimate. Shareable line via --share.
Efficient Literature Survey
Claude Code skill for efficient literature survey with 99%+ token savings
Claude Tools Dashboard
Live token savings wallboard for Claude Code tools (RTK, Headroom, jCodeMunch, jDocMunch)
Ashlr Benchmark
Run the ashlr-plugin token-savings benchmark against the current project.
Related Agents
Design Guardian
Accountable design decider and applier. Reviews real UI evidence, rejects token drift and generic AI styling,
Impact Analyst
Calculates time savings, ROI, and productivity improvements from interventions
Bess Analyst
Analyze BESS issues, debug problems, and explain system behavior. Use when investigating savings calculations,