aSkills.dev logo
aSkills.devRegistry
Submit Skill
Verified SOP & TutorialAudited Score 98/100

The Complete Production SOP: How to Deploy and Run AppArmor Linux Security Module Profiles in Autonomous AI Agents

Step-by-step SOP to install and run AppArmor Linux Security Module Profiles in Codex, Claude Code, and Cursor. Confine binaries with verified rules.

View 1-Click Spec
Target Runtimes:codexclaude-codecursorwindsurf
5 min read • Updated for 2026

Technical Overview & Architectural Context

Autonomous coding agents executing system-level hardening often struggle with Linux Security Modules (LSM) like AppArmor. Left unguided, agents frequently hallucinate overly broad wildcards (e.g., /** rwk), misplace include abstractions (#include <tunables/global>), or omit critical capability delegations (capability net_bind_service), rendering profiles either porous or completely broken. In containerized or microservice environments, an agent might prematurely enforce a profile without parsing audit.log denials, causing unhandled SIGKILL terminations on essential daemons. The apparmor-profile-enforcement skill enforces strict procedural boundaries for profile authoring and lifecycle management. It forces the agent to follow a deterministic pipeline: profile scaffolding via aa-genprof, granular file path scoping (distinguishing execute flags like px, cx, and ix), complain-mode log learning via aa-logprof, and finally rigid enforcement via aa-enforce. This ensures system services run under verifiable least-privilege mandatory access controls without accidental production downtime.

Prerequisites & Environment Requirements

Linux host or VM running kernel 5.4+ with AppArmor LSM enabled (cat /sys/module/apparmor/parameters/enabled returning Y)
apparmor-utils, auditd, and apparmor-profiles installed on the target environment
Codex Desktop, Claude Code CLI, Cursor, or Windsurf IDE with workspace access
Root or sudoer capability on the target host for kernel profile loading and enforcing

Verified Search Intents & Long-Tail Queries Solved

SEO Audit Score: 98/100

This operational tutorial is engineered to provide deterministic solutions for these developer search queries:

#install apparmor-profile-enforcement in cursor ide#claude code apparmor-profile-enforcement rules#codex desktop apparmor-profile-enforcement setup tutorial#how to debug apparmor denials with aa-logprof in claude code#how to fix apparmor-profile-enforcement agent execution error#mandatory access control apparmor profiles for docker containers#apparmor-profile-enforcement sandbox permissions and security score#apparmor-profile-enforcement prompt examples and triggers

Step-by-Step Execution Playbook

1

Local Environment Setup & SKILL.md Mounting

Create the local skill manifest directory and inject verified AppArmor procedural rules.

Autonomous runtimes such as Claude Code and Codex Desktop discover operational skills via workspace or global manifest paths (e.g., .agents/skills/ or .claude/skills/). By placing the SKILL.md file in the project or agent directory, the runtime dynamically parses the frontmatter triggers ('create apparmor profile', 'enforce apparmor profile', 'debug apparmor denials') and mounts the execution rules into its decision graph.

Terminal Command
mkdir -p ~/.agents/skills/apparmor-profile-enforcement && cat << 'EOF' > ~/.agents/skills/apparmor-profile-enforcement/SKILL.md --- name: "apparmor-profile-enforcement" description: "Confine Linux applications and microservices to strict least-privilege profiles with AppArmor." metadata: author: "AppArmor" version: "1.0.0" runtime_compatibility: - "codex" - "claude-code" - "cursor" - "windsurf" triggers: - "create apparmor profile" - "enforce apparmor profile" - "debug apparmor denials" permissions: fs_read: true fs_write: "workspace" shell_exec: false net_access: false --- # AppArmor Linux Security Module Profiles Authors and enforces AppArmor profiles (/etc/apparmor.d): aa-genprof profiling, file access restrictions (r, w, m, k), network family permissions, and aa-enforce activation. ## Execution SOP & Rules 1. Scaffold profile generation using aa-genprof /path/to/binary or author declarative abstractions in /etc/apparmor.d/. 2. Define file access permissions explicitly with distinct access modes (r, w, m, k, px, ix). 3. Test in complain mode (aa-complain) and inspect audit logs via aa-logprof. 4. Transition profile to enforce mode using aa-enforce /etc/apparmor.d/profile. 5. Verify active enforcement status using aa-status command. EOF
SKILL.md
---
name: "apparmor-profile-enforcement"
description: "Confine Linux applications and microservices to strict least-privilege profiles with AppArmor."
metadata:
  author: "AppArmor"
  version: "1.0.0"
  runtime_compatibility:
    - "codex"
    - "claude-code"
    - "cursor"
    - "windsurf"
triggers:
  - "create apparmor profile"
  - "enforce apparmor profile"
  - "debug apparmor denials"
permissions:
  fs_read: true
  fs_write: "workspace"
  shell_exec: false
  net_access: false
---
Production Pro-Tip

Keep the SKILL.md file free of raw log dumps. Have the agent parse output summaries using aa-status --json to minimize token consumption during profiling analysis.

Sandbox Gotcha

AppArmor profile parsing tools (apparmor_parser) require root permissions to load profiles into the Linux kernel (/sys/kernel/security/apparmor). Ensure the agent generates files in the workspace before requesting human-in-the-loop sudo elevation.

2

Environment Secrets & Dependency Validation

Verify host LSM capabilities and required AppArmor utilities without granting broad sudo access.

Before the agent creates profiles, it must verify that AppArmor is active on the host and that required binaries (aa-status, apparmor_parser, aa-complain, aa-enforce) are present. Because the skill has shell_exec set to false for safe execution, the agent delegates verification commands to the developer or executes them within a pre-approved, scoped execution container.

Terminal Command
which aa-status aa-enforce aa-complain apparmor_parser && cat /sys/module/apparmor/parameters/enabled
Production Pro-Tip

When writing profiles for containers (e.g., Docker or Podman), specify profile names using the 'profile <name> flags=(attach_disconnected,mediate_deleted)' syntax to avoid runtime bind-mount access rejections.

Sandbox Gotcha

Do not execute 'aa-enforce' directly within an unprivileged unshare namespace or unprivileged Docker container without CAP_MAC_ADMIN; the kernel will reject policy insertion with EPERM.

3

Intent Matching & Trigger Invocation

Activate the AppArmor profiling workflow using deterministic triggers.

Prompt the agent using unambiguous trigger phrases matching the frontmatter triggers ('create apparmor profile', 'enforce apparmor profile', or 'debug apparmor denials'). The runtime matches these patterns and forces the agent to follow the prescribed SOP instead of emitting generic bash scripts.

Production Pro-Tip

Specify the full absolute binary target path and known runtime dependencies (e.g., systemd service files, log directories) in your trigger prompt to avoid iterative file-path discovery cycles.

Sandbox Gotcha

Avoid mixing multiple triggers in a single prompt (e.g., 'create and enforce and debug'). Handle profiling in distinct lifecycle stages: scaffold -> complain -> audit -> enforce.

4

Verification, Linting & Output Quality Gate

Perform static syntax validation and kernel-level checks before committing AppArmor profiles.

The skill mandates running local linting and parser validation prior to moving profiles into /etc/apparmor.d/. The apparmor_parser utility with the -d (debug/dump) or -Q (skip kernel loading) flags allows syntax verification without requiring root escalation or disrupting host security state.

Terminal Command
apparmor_parser -Q -v ./profiles/usr.bin.custom-microservice
Production Pro-Tip

Use 'apparmor_parser -r -W <profile>' during CI/CD to pre-compile binary policy caches and catch malformed regular expressions prior to server deployment.

Sandbox Gotcha

Missing include files (e.g., #include <abstractions/base>) will cause apparmor_parser to fail with file not found. Ensure required AppArmor abstraction packages are present in the build environment.

Real-World Scenario: Hardening a Node.js API Gateway Microservice with AppArmor

Tested Trigger Prompt (Feed directly into Codex, Claude, or Cursor):

“We need to create apparmor profile for our Node.js microservice located at /usr/local/bin/api-gateway. It reads configs from /etc/gateway/*.json, logs to /var/log/gateway/*.log, binds to TCP port 8080, and needs read-only access to standard Node.js dynamic libraries. Generate the profile, transition it to enforce mode, and verify enforcement.”

Deterministic Agent Verification Pipeline:
Agent analyzes the provided path and imports standard AppArmor base abstractions (#include <abstractions/base>, #include <abstractions/nameservice>, #include <abstractions/node>).
Agent authors the confined profile '/etc/apparmor.d/usr.local.bin.api-gateway' specifying exact read access for /etc/gateway/*.json, append/write access for /var/log/gateway/*.log, and 'network inet stream' for TCP port binding.
Agent instructs the user to validate the profile syntax using 'apparmor_parser -Q ./usr.local.bin.api-gateway'.
Agent provides the step-by-step transition commands: initial load via 'sudo apparmor_parser -r', testing via 'sudo aa-complain', log verification via 'sudo aa-logprof', and final transition using 'sudo aa-enforce /etc/apparmor.d/usr.local.bin.api-gateway'.
Agent confirms active confinement by verifying the output of 'sudo aa-status | grep usr.local.bin.api-gateway'.

Runtime Troubleshooting & Failure Modes

Error: AppArmor parser error for /etc/apparmor.d/usr.local.bin.service at line 14: syntax error, unexpected TOK_ID, expecting TOK_MODE

Root Cause: Missing permission mode flags on a path rule, or missing trailing comma at the end of a profile entry.

Resolution Command / Fix:
# Fix line 14 by appending required access modes (r, w, m, k) and trailing comma:
# Incorrect: /var/log/service.log
# Correct:
/var/log/service.log rw,
Error: audit: type=1400 audit(1710934120.412:89): apparmor="DENIED" operation="exec" profile="/usr/local/bin/api-gateway" name="/bin/dash" pid=4124 comm="node" requested_mask="x" denied_mask="x"

Root Cause: The application attempted to spawn a child process (/bin/dash via child_process.exec) without explicit transition rules (px, cx, ix, or ux) defined in the AppArmor profile.

Resolution Command / Fix:
# If inheritance without sanitation is intended, add ix to the profile:
/bin/dash rix,
/bin/sh rix,
# Then reload the profile into the kernel:
sudo apparmor_parser -r /etc/apparmor.d/usr.local.bin.api-gateway

Frequently Asked Questions (AppArmor Linux Security Module Profiles SOP)

Q:Can I use AppArmor Linux Security Module Profiles across Codex, Claude Code, and Cursor simultaneously?

Yes. The apparmor-profile-enforcement skill conforms to the universal SKILL.md specification. When cloned into your shared ~/.agents/skills or project-level .agents directory, Codex Desktop, Claude Code CLI, Cursor, and Windsurf can all execute the exact same profile-generation SOP.

Q:How does AppArmor Linux Security Module Profiles prevent context window overflow during long agent sessions?

The skill enforces structured SOP boundaries: it directs the agent to generate standard abstractions (#include <abstractions/...>) and validate profile definitions using deterministic CLI flags (apparmor_parser -Q) instead of feeding thousands of raw dmesg or auditd logs directly into the LLM context window.

Q:Does this skill require escalated root or container permissions?

No. The skill itself has a 98/100 security score with shell_exec and net_access disabled. The agent drafts and validates AppArmor profiles inside your isolated workspace. Applying the profile to the live kernel (/etc/apparmor.d/ and apparmor_parser -r) is safely decoupled and run by the developer via explicit sudo commands.

Ready to equip your autonomous agent with AppArmor Linux Security Module Profiles?

Copy the complete audited SKILL.md specification or download the 1-click zip bundle directly from our verified registry.

Get 1-Click Install Spec