- Language
- TypeScript
- Last commit
- 56m ago
- Forks
- 1
- Open issues
- 0
- Registry update
- Oct 13, 2026
Connect
remote + npm$claude mcp add --transport http aegis-skills https://aegis-skills.vercel.app/api/mcpHeaders
Nothing to configure — it works as is.
Security review
Review passedReviewed 42m ago.
- tools
- 3 tools scanned
- metadata
- scanned
- packages
- 1 checked
No findings.
Tools
3list_skills
List available defensive security skills, optionally filtered by family or a free-text query (matched against name, description, tags and frameworks).
get_skill
Return a skill's description, frameworks, tags, context and ordered phase ids.
get_skill_phase
Load the methodology content for one phase of a skill. The phase is given as a phase id or as a 0-based index among authored phases.
aegis-skills
24 defensive cybersecurity skills in the open Agent Skills format, with a native installer and a read-only MCP server.
Install
npx aegis-skills install --platform claude # or codex, gemini, cursor, vscode, antigravity, all
npx aegis-skills install --platform claude --project
Skills go into each tool's own skills folder and the aegis MCP server is registered in its config (the file is backed up first). After an upgrade, installed skills refresh automatically; files you edited are kept as .bak. Opt out with AEGIS_NO_AUTO_REFRESH=1.
Claude Code plugin: /plugin marketplace add drupadsachania/aegis-skills, then /plugin install aegis@aegis-skills.
Commands
install (alias init), uninstall, update, list, mcp. Run aegis --help.
MCP
npx aegis-skills mcp serves list_skills, get_skill and get_skill_phase on stdio. Hosted: https://aegis-skills.vercel.app/api/mcp.
What this plugin runs, sends and fetches
- Claude Code plugin (
.mcp.json): connects to one remote MCP server,https://aegis-skills.vercel.app/api/mcp. It runs no local command. The three tools (list_skills,get_skill,get_skill_phase) are read-only and return fixed public skill text. What is sent is the tool call itself (a skill name, a phase or a search word) and the usual request metadata such as your IP address. The server keeps only an in-memory request counter, makes no model calls and stores nothing. The skills themselves are plain Markdown that your assistant reads locally. npx aegis-skills(the installer and stdio server): runs the pinned npm packageaegis-skills.installwrites skill files into your tool's skills folder and adds anaegisentry to its MCP config (backed up first). A ledger of what it installed is kept under~/.aegis(override withAEGIS_HOME).mcpserves the same three read-only tools from the bundled skills on stdio. Neither command makes a network request of its own beyond npm downloading the package.- No telemetry, no analytics and no credentials. Nothing is read from your conversations, memory or uploaded files.
Every skill is for defensive use on systems you own or are explicitly authorized to assess.
Licence: Apache-2.0 OR MIT.
From the server's repository — third-party content, read it before running anything it suggests.
More in Security
depwire
io.github.atef-ataya
Dependency graph + 24 MCP tools. Impact analysis, simulation, security, agent coordination
SUNGLASSES
io.github.sunglasses-dev
Input firewall for AI agents: scan text and files before an agent reads them. Local, no network.
CanaryGuard 402
io.github.ak10082247-max
Deterministic prompt-injection detection & sanitization for AI agents via x402 on Base.
zn-gate
io.github.usezn
Local security layer for coding agents: prompt-injection checks, secret redaction, MCP proxy.
attack-surface-mcp-server
io.github.cyanheads
Passive external attack-surface mapping: CT subdomains, DNS, TLS, HTTP posture, RDAP/WHOIS, Shodan.
MCP Dataverse
io.github.codeurali
Microsoft Dataverse MCP server: CRUD, FetchXML, metadata, schema, audit, batch and security roles.
Listing from the official MCP Registry · repository facts from GitHub, 44m ago.