infra-agent-skills

← All skills

Infrastructure Plan Execution

infrastructure-plan-execution

Prepare, review, and apply a declarative infrastructure plan, or plan its removal, then inspect the resulting recorded state and outputs and adopt an existing resource under management. Use when a configuration directory must be initialized, validated, planned, applied, or destroyed through a reviewed saved plan. Covers HashiCorp Terraform CLI. Assumes the CLI, an existing configuration, state backend access, and provider credentials are already available; this skill does not install tools or author configuration. Does not cover writing resource definitions, provider-specific resource design, or approval workflows.

To install (framework-agnostic: plain markdown, nothing to compile or pip install. Copy the skill folder to wherever your agent framework looks for skills, e.g. for Claude Code):
git clone https://github.com/itential/infra-agent-skills.git
cp -r infra-agent-skills/skills/<skill-name> ~/.claude/skills/

Infrastructure Plan Execution

Prepares a configuration, produces a reviewed plan, applies or destroys exactly that plan, and reads back recorded state, using the platform’s own documented procedure and result vocabulary.

When to use this skill

  • A configuration directory must be initialized or validated before a change.
  • A plan must be produced, reviewed, and then applied, or a removal plan produced and executed.
  • Recorded resource addresses or outputs must be listed.
  • An existing remote object must be bound to a resource address in the configuration.
  • Not for writing or redesigning the configuration, installing the tool, or managing credentials.

Universal concepts (apply regardless of platform)

  • A plan describes proposed changes and is separate from applying it; applying can change or delete real resources.
  • Where a platform documents a saved plan, applying that exact artifact is the reviewed path, and a failed or unreviewed plan is not eligible.
  • Where documented, a failed apply is not rolled back automatically, so a failure can leave the infrastructure partially changed.
  • Where documented, plan artifacts and state can hold sensitive values that terminal output hides.
  • No wire-level standard governs this capability; claims here are hedged accordingly.

Procedure

  1. Identify the platform from the request or the platform argument. If it is not in references/INDEX.md, stop; do not improvise a procedure.
  2. Follow references/<platform>/procedure.md exactly, including its confirmation gates.
  3. Report the outcome in that procedure’s own result vocabulary, preserving partial or incomplete results.

Known limitations

  • Does not author configuration, choose providers, or design resources.
  • Platform coverage is exactly the platforms in references/INDEX.md.

Contributing to this skill

Change this router only to correct shared concepts or add a pointer to a newly supported platform. Platform-specific operations, errors, and bindings belong under that platform’s references/ directory.

Tools

A local command-line client for the platform, run in the configuration directory. Each procedure lists the operations it uses.

Verification checklist

  • Platform confirmed against references/INDEX.md
  • That platform’s procedure.md followed, including confirmation gates
  • Apply or destroy used the exact plan the operator reviewed
  • Result reported in the platform’s own vocabulary, with partial failure stated