Atmos is an infrastructure runtime that coordinates tools such as Terraform, OpenTofu, Kubernetes, Helm, Packer, Ansible, and containers through consistent commands and configuration. It is for teams running cloud infrastructure on laptops, in CI, or through AI agents across environments and regions. Its catalogue entries provide skills, agents, commands, and other add-ons for Atmos workflows.
Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/cloudposse/atmosnpx agentmods add skills/cloudposse/atmos/atmos-policyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cloudposse/atmos/atmos-policy)<a href="https://agentmods.dev/skills/cloudposse/atmos/atmos-policy"><img src="https://agentmods.dev/badge/skills/cloudposse/atmos/atmos-policy/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cloudposse/atmos/atmos-policy"><img src="https://agentmods.dev/badge/skills/cloudposse/atmos/atmos-policy.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00051 | $0.01741 |
| Opus 5.5 | $0.00020 | $0.00696 |
| Sonnet 5.5 | $0.00010 | $0.00348 |
| Haiku 4.5 | $0.00005 | $0.00174 |
Grade A, and why
atmos-policy scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 15d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Atmos Policy
Use this skill for the experimental policy engine and reusable library, not for general validation
or ordinary lifecycle-hook configuration. Use atmos-hooks for non-policy hook definitions and
atmos-validation for existing JSON Schema/OPA settings.validation behavior.
Establish capabilities before authoring
Check the installed Atmos version/help and the matching source/docs. Do not infer a command exists
from the PRD: atmos policy check/test/explain is proposed, not implemented in this branch.
The current opt-in is policy.d/ at the project root that atmos.yaml discovery selects (the working directory, then the Git root, then the nearest ancestor with atmos.yaml), overridden by ATMOS_POLICY_DIR, containing strict atmos/v1
kind: Policy documents. Inline CEL and restricted Starlark run. A Starlark predicate is a
function body over detached, read-only JSON data that must return a strict boolean; it has no
automation host, module loader, filesystem, network, or logging API. Rego is not available
through this loader. Existing Rego validation is separate and remains supported. Do not migrate legacy configuration to proposed policy:
attachments yet.
Use central match selectors for instance attachments: literal labels (all), tags (any by default,
tag_mode: all supported), exact component names, stacks, and component types. These use resolved
metadata from existing inheritance, not inline policy inheritance. Protect the governing directory.
Distinguish input kind, component type, and CI provider:
terraform.plan: native JSON extracted privately from a saved binary plan. Evaluated after successful plan, including no-change plans, and again before applying the evaluated private copy. Implicit apply/deploy and direct destroy are rejected when plan policies are enabled.operation: lifecycle event, component type/name, stack, command, status, exit code, dry-run. Current integration uses existing Terraform, Helmfile, Helm, and Kubernetes hooks. Do not claim enforcement for Packer, Ansible, containers, emulators, or custom components yet.plugin(on: plugin): native plugin registries, packages, and local builds (origin,path,profiles,operation) checked before install, activation, test, publish, and sign. Seeatmos-pluginsfor inputs and recipes.- GitHub Actions context:
ci.provider == 'github-actions',ci.schema == 'github-actions/v1'. Metadata retains allowlisted environment variable names underci.metadata.data; native event fields remain underci.event.data. No tokens or arbitrary environment variables are copied. Events are triggering evidence, not current reviews. Missing required data blocks. atmos.affected: deny rules run in shared affected analysis before output filtering/scheduling. Inputs include files, instances, component/stack groups, and structured before/after fields. Collection requires clean worktrees; static inventories use the active, protected discovery configuration on both revisions. This is not independent historical discovery or a transitive execution graph. Missing imports and unsupported inventory types block. Inspect before/after explicitly; affected rules do not accept single-instancematchor dispatch lifecycle hooks.git.change, function/source, and approval input types have core contracts or design examples, not complete CLI collectors/enforcement. Affected is explicit; ordinary lifecycle and saved-plan rules do not depend on--affected.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 15d ago Changed · +2 lines 70a58fe82431
- 24d ago First seen · 117 lines · 51 tokens per session scan A cd0385a5937d
atmos-policy is a skill published in the GitHub repository cloudposse/atmos (1,419 stars, last pushed today), licensed Apache-2.0. It adds 51 tokens to every session and 1,741 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-20.
atmos-policy compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| atmos-policycloudposse/atmos | Skill | 1.4k | today | Apache-2.0 | A | 51 | Claude Code |
| gke-basicsskillmds/skillmd | Skill | 142 | 4d ago | MIT | A | 4 | Claude Code |
| clum-mcptddh/clum | Skill | 6 | 19d ago | MIT | F | 19 | Claude Code, Codex |
| x-netsx-cmd/x-cmd | Skill | 4.8k | 13d ago | Apache-2.0 | A | 69 | Claude Code, Codex |
| x-osvx-cmd/x-cmd | Skill | 4.8k | 13d ago | Apache-2.0 | A | 72 | Claude Code, Codex |
| x-uptimex-cmd/x-cmd | Skill | 4.8k | 13d ago | Apache-2.0 | A | 67 | Claude Code, Codex |
| tsvx-cmd/x-cmd | Skill | 4.8k | 13d ago | Apache-2.0 | A | 47 | Claude Code, Codex |