infisical/ai-skills

infisical-pam

Guide for Infisical Privileged Access Manager (PAM) — brokering human and AI-agent access to databases, servers, Kubernetes clusters, and cloud accounts without the connecting party ever seeing a credential, with full session recording and audit.

View source

How to use

  1. Copy the installation command and review the source package before installing it in your own agent environment.
  2. Check the source document for required tools, credentials and supported hosts.
  3. Describe your input and desired deliverable; review the output against the source material.
npx skills add https://github.com/infisical/ai-skills --skill infisical-pam

Before you start

  • Installation does not include a model subscription, API allowance or permission to access your files.
  • Repository stars and installation counts describe activity, not an independent quality or security test.
Original skill document

Rendered from the source repository. Headings, examples, code, tables, links, and referenced images are preserved.

Infisical Privileged Access Manager (PAM) Guide

You are a setup assistant helping users broker privileged access to infrastructure through Infisical PAM.

The core property: the connecting party never sees the credential. A person or an AI agent connects through Infisical, Infisical holds the real credential and proxies the connection, and every query or command is recorded.

PAM is a separate Infisical product with its own project type (pam). Enterprise feature.

Not this skill

The boundary is who or what is connecting, and whether they see a credential:

If the user wants...Use
A person or agent to reach a database/server without holding a credential, with recordingthis skill
An application to fetch a credential and connect itselfinfisical-dynamic-secrets or infisical-setup
An ephemeral credential handed to a workloadinfisical-dynamic-secrets
An existing credential rotated on a schedule into secretsinfisical-secret-rotation
Approval workflows for secrets, not infrastructureinfisical-access-control
To deploy the Gateway that PAM proxies throughinfisical-gateway

The sharpest distinction is PAM versus dynamic secrets. Both give time-bound access, but:

  • Dynamic secrets hand a real credential to the consumer. Good for applications and CI, which need to connect themselves.
  • PAM never hands over a credential. Infisical proxies the session. Good for humans and AI agents, where you want recording and the ability to terminate mid-session.

If the user says "I want engineers to query production without knowing the password," that is PAM. If they say "my service needs a database credential," that is dynamic secrets.

The four components

Get this model right and the rest follows:

ComponentWhat it is
AccountOne database, server, cluster, or cloud account. Holds connection details and credentials
FolderA grouping of accounts by who needs access. Permissions are set here
Account TemplateThe policy applied when connecting: max session duration, reason required, MFA required. Type-specific
MembershipAssigns a user, group, or machine identity a role on a folder or account

The rule to remember: templates define rules, folders define access.

  • Accounts in the same folder can use different templates — same team, different rules
  • Accounts in different folders can share a template — same rules, different teams
  • Folder memberships cascade to every account inside; assign directly on an account only for exceptions

Organize folders by who needs access, not by technology. backend-team is a better folder than postgres-databases, because permissions live at the folder level.

Roles

RoleCapabilities
AdminFull control — accounts, folders, sessions, memberships
ConnectorLaunch sessions and connect to accounts
AuditorView audit logs and session recordings

Machine identities hold these same roles. Connector is the usual choice for an unattended AI agent.

There is also a product-level role split (admin, member) governing who can create Accounts and Folders at all.

How to use this skill

  1. Confirm PAM is right — the connecting party must not hold a credential (see the table above)
  2. Check the Gateway story — PAM proxies through a Gateway; internal targets need one
  3. Model folders around teams, not technologies
  4. Create accounts with connection details and credentials
  5. Pick or create templates for session policy
  6. Assign memberships at the folder level
  7. Decide on just-in-time access — standing Connector access, or access requests with approval
  8. Confirm recording and auditing meet the requirement

Reference files

FileWhen to read
references/accounts-and-access.mdAll 13 account types, per-type auth options, folders, templates, memberships, access requests, discovery, dependencies
references/sessions-and-agents.mdSession lifecycle and states, recording, termination, web vs CLI access, agentic access for AI agents

Guiding principles

  • Never route a credential to the user. The whole value of PAM is that they do not get one. If a user asks how to retrieve the underlying password, explain that PAM deliberately does not expose it.
  • PAM needs a Gateway for anything internal. Connections are proxied. See infisical-gateway.
  • Folders are the permission boundary. Someone with access to a folder reaches every account in it. That makes folder design a security decision, not an organizational nicety.
  • Recording is real-time. Recordings can be reviewed while a session is still active, and an admin can terminate mid-session. Mention this for incident-response scenarios.
  • Require a reason and MFA on production templates. They are template settings, cheap to enable, and they are what makes the audit trail meaningful after the fact.
  • Agentic access is the right answer for AI agents. infisical pam agentic access -- <agent> sandboxes the agent and brokers its access; do not suggest putting a connection string in an agent's environment.
  • Session TTLs are ceilings, not targets. Set the maximum session duration to the shortest thing that lets real work finish.
from this repository

More skills

All skills
infisical
Community

infisical-access-control

Guide for Infisical access control and governance — org and project roles, custom roles with subject/action permissions, granular secret permissions that separate describing a secret from reading its value, attribute-based access control (ABAC) driven by user metadata and machine identity attributes, additional privileges, assume-privilege, temporary (time-bound) access, access requests, change-approval and access-approval policies, and audit logs including streaming to Splunk, Datadog, Azure, C

1h installs
1
GitHub stars
41
Updated
Nov 11
infisical
Community

infisical-agent

Guide for configuring the Infisical Agent — a client daemon that manages token lifecycle and renders secrets via Go templates without modifying application code. Covers the full YAML config format, all 6 agent auth methods (universal-auth, kubernetes, aws-iam, azure, gcp-id-token, gcp-iam), sinks, template functions (listSecrets, listSecretsByProjectSlug, getSecretByName, dynamicSecret), polling, on-change commands, retry strategy, and persistent caching. Use this skill when someone asks about:

1h installs
1
GitHub stars
41
Updated
Nov 11
infisical
Community

infisical-api

Interact with the Infisical REST API to manage secrets, projects, environments, machine identities, and more. Supports secret CRUD operations, machine identity authentication, pagination, and rate limiting on cloud deployments. Not for the CLI/SDKs (infisical-setup), KMS crypto endpoints (infisical-kms), certificate endpoints (infisical-pki), or human SSO login (infisical-sso).

1h installs
1
GitHub stars
41
Updated
Nov 11
infisical
Community

infisical-app-connections

Guide for creating and managing Infisical App Connections — the reusable authenticated links between Infisical and third-party services that Secret Syncs, Secret Rotations, PKI certificate authorities, PKI Syncs, and Secret Scanning data sources all depend on. Covers all 83 connection types and their exact auth methods (AWS assume-role vs access-key, GitHub OAuth vs GitHub App vs PAT, Azure OAuth vs client-secret vs certificate, GCP service account impersonation, and more), platform-managed cred

1h installs
1
GitHub stars
41
Updated
Nov 11