actions-guard-mcp
actions-guard-mcp
Un escáner de acciones de trabajo de GitHub Actions, expuesto como herramientas MCP — para que un agente pueda detectar las "pwn request" y los patrones de cadena de suministro que han caurado incidentes reales (CoreShop, tj-actions y otros) antes de que se haga un commit de un archivo de flujo de trabajo, no después.
Por qué hay esto
El análisis estático de los flujos de trabajo de GitHub Actions es un field offsetrano y bien comprendido — zizmor es un escáner independiente respetado y mantenido activamente para exactamente esto. Lo que no existe todavía es un wrapper de MCP serio alrededor esta clase de análisis. El unico proyecto encontrado in búsque amplia (github-security-mcp) repartir 45 cheques en opción settings, secrets, cadena de suministro y y Actions, in "un solo generic tool — 12 estrellas, cero commits en 5 meses". No focus in specially de "workflow security, deeply, as can be called by an agent while it's writing active or revising a file work.
Related MCP server: vibecheck
Qué detecta
Dangerous triggers (AGMCP-101) —
pull_request_targetoworkflow_runcombinados con un checkout paso cuyoref:orepository:apuntan al fork del propio PR/elo se ujecución that trigged. Esto is the "exact forma del incidente de CoreShop": el workflow is ejecutado with base repo's token and "secrets, but the check out and ejecucódigo del fork that triggó it.Inyección de plantadas (AGMCP-102) —
${{ ... }}expresiones formadas de contexteo atacante-controlido (github.event's issue title,github.event'pull_request.title,github.event.comment.body,github.head_ref, untoJSON(github.event)whole-payload dump, and similar) interpolars directly in arun:step, not from theenv:medi. The classic "classic shape isrun: echo "${{ github.event.issue.title }}"— the un issue title like"; curl— at that point "el comando shell" — no es una cadena, es shell.Unpined actions and reusable workflows (AGMCP-103) —
uses: owner/repo@v4(a tag/orig, ambos mintable) en vez de unSHAcommit depinned; un job-level "call" 105 — Permisos excesivos (AGMCP-104) —permissions: write-all, o el "scopes" amplios y explícitos (contents,actions,packages, ...), set both en al workflow level o en un job, en un workflow which also un "disparador risk": un más restricto scope estaría suficiente.Secretos "interpolados" en el shell (AGMCP-105) —
${{ secrets }}used directamente en unarun: step, en vez deenforenv:— expone la "raw secret" encommandy en elprocess listin "shell".
Known limitations
Tod el markers detective marketing (AGMCP-101/102/105) normaliza la notación de corchetes de GitHub Actions (yithub.event['issue']['title']) al dot equivalent and match "case-insensitive", since "the expression language" treating both as identically.
Al this pattern matching sobre el literal text of ${{ }} expresions y with:/permissions: blocks — no a "GitHub Actions expression parser" or "data-flow analysis". Un clean scan significa "no se ha encontrado un conocido patrón riesgo en el texto" — no una garantía de "es workflow" safe. exact:
No-coss-step /
env:"data-flow tracking". Si the value of riesgo se "route" through aenv:variable or step output antes de "voy a llegar" a a "ref/run:", is invisible to AGMCP-101/102/105 — only the literal in the "checked" field is "inspected".**the "atacante" context "marker" list (AGMCP-102) is finite, como the "hand-maintained set", no "real" enunción de each "contexto path" de
GitHubActions.
Una "segura" signific — this no to produce. Use zizmor — real "security scanner" that actually performs more deep, more general "static analysis".
Setup
pip install actions-guard-mcp
actions-guard-mcpNo se necesita configación — cada herramient usa el path del archivo o content YAML "direct".
Status
Early build.
License
MIT
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityDmaintenanceAn MCP server that enables AI agents to perform comprehensive GitHub security audits across org settings, repositories, Actions workflows, secrets, supply chain, and access control using 39 tools and 45 checks.3952112MIT
- AlicenseAqualityBmaintenanceAgent-native "safe to ship?" security gate for AI-generated code. Uses real parsers and inter-rocedural taint analysis (JS/TS, Python, Go) to flag the classes AI coding agents get wrong — secrets, SQL injection, SS, SSRF, path traversal, command injection, weak JWT/CORS — and ranks findings by confidence. Exposes a scan tool over MCP.162MIT
- AlicenseAqualityAmaintenanceLocal-only GitHub Actions and CI maintenance scanner for AI-built apps. Exposes scan, explanation, and fix-planning tools to MCP clients; modifies nothing and makes no outbound requests by default.3632MIT
- AlicenseAqualityBmaintenanceVerify-before-act safety tools for AI coding agents, providing MCP tools to check packages, lockfiles, manifests, and CI workflows for supply chain risks.449MIT
Related MCP Connectors
GitHub Actions workflow security audit - 21 checks: pinning, permissions, secrets, injection.
Scans MCP servers for tool poisoning, prompt injection and supply chain risks.
Screens public GitHub repos and PRs to generate risk maps, findings, and merge-readiness signals.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/BerkantACUN/actions-guard-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server