Works with the AI Agents you already use
Security Aware Repository Indexer
Index any codebase into token-efficient AI context, carry session memory across restarts, and audit hidden MCP servers, hooks, and npm supply-chain entries.
Free
Security Aware Repository Indexer
Example session with this skill installed
I just cloned a fintech monorepo at ~/work/payments-platform and my agent keeps blowing the context window trying to read it. Also, last week a colleague's machine got hit by an npm postinstall that registered a rogue MCP server, so I want every AI-related config on this box checked before we start. Can you use Verdant Scan to build a compact map of the repo - languages, manifests, entry points, recent changes - then run the full audit for lifecycle hooks, remote-mutable references, and hidden agent state, and save the working context as a named session I can restore later?
- Read your context and instructions
- Compiled the security aware repository
Summary
Verdant Scan indexed ~/work/payments-platform (1,284 files) and audited every AI-related config on this machine; two untrusted entries were removed. Project map: rust 68% / python 22% / markdown 10%, manifests Cargo.toml + pyproject.toml, entry points src/main.rs and workers/charges.py.
| Check | Before | After |
|---|---|---|
| Files mapped | raw tree, 14k lines | compact summary, 96 lines |
| MCP servers registered | 5 | 3 |
| Lifecycle hooks active | 4 | 1 (inspected) |
| Remote-mutable refs | 2 | 0 |
| Session state | lost on restart | saved as checkout-bug |
Next steps
- Delete the untrusted npm package before re-running the audit, or its postinstall will recreate the deleted entries
- Re-run
vs audit --allweekly; feedvs render --mode compact --max-lines 200 --summaryinto agent prompts instead of raw trees - Resume with
vs session restore --label checkout-bug; memory decision anchored (httpOnly cookies for session tokens) - Keep the dashboard bound to 127.0.0.1; add auth and read-only mode before any remote use
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Large codebases overwhelm LLM context windows, leading to hallucinations and high token costs. Developers also face hidden security risks from untrusted npm packages or editor plugins that register silent AI hooks and MCP servers.
What it does
- Generates token-efficient, compact project maps and summaries for AI assistants.
- Audits local environments for suspicious MCP servers and hidden agent lifecycle hooks.
- Maintains persistent session memory and searchable notes across IDE restarts.
- Provides a rapid directory visualizer with filtering, sorting, and regex search.
- Exposes a Model Context Protocol (MCP) server for direct agent interaction.
Frameworks & tools
Rust, Cargo, Model Context Protocol (MCP), Node.js/npm, and major coding assistants like Claude Desktop and Cursor.
Why this beats prompting it yourself
Raw directory listings and file dumps waste tokens and lack structural context. This skill uses a daemon-backed cache and compact rendering modes specifically tuned for agent ingestion, saving time and money while improving reasoning accuracy.
Use cases
- Onboarding an AI assistant to a massive legacy repository without hitting token limits.
- Auditing a machine for unauthorized AI integrations after installing new dev tools.
- Resuming complex refactoring tasks by restoring saved agent session memory.
- Creating structured architecture maps and changelogs for technical documentation.
Known limitations
Performance may degrade on Windows when scanning projects mounted via WSL /mnt/c. It identifies suspicious patterns but does not provide a sandbox for executing untrusted code.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Scanning Windows-mounted WSL paths is slower than native filesystems. The audit tool detects suspicious patterns but is not a sandbox or a replacement for full code signing policies.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
3 uses
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 2 months ago
- Free with an Agensi account