ClaudeSuperPower

idmp-shared

Skill

Shared IDMP CLI rules for choosing element mode vs template mode, resolving business-root rootElementId, acknowledging risky writes, and verifying results.

Install

git clone https://github.com/taosdata/agent-skills.git ~/.claude/skills/idmp-shared

What is idmp-shared?

Shared IDMP CLI rules for choosing element mode vs template mode, resolving business-root rootElementId, acknowledging risky writes, and verifying results.

What this can do

Capabilities declared in this component's own frontmatter — not inferred.

Inherit all session tools

Declares no tool restrictions — inherits every session tool

~39 tokens of context used while enabled, before you invoke anything

Documentation

README · ~5 min read

Shared IDMP CLI rules

🛑 Destructive op confirmation (mandatory)

Before executing any DELETE, non-readonly POST / PUT / PATCH, --ack-risk command, or local destructive helper (config remove / profile remove / auth logout --clear), the agent MUST print the 4 items below and then wait for an explicit user yes before running the command:

  1. Target — resource type + id + name + owning element / template / profile.
  2. Action — HTTP method + path + schemaPath + param summary (mask secrets).
  3. Reason — why this operation is needed (trace it back to the user's request; if it is a rollback of a failed previous step, say so).
  4. Blast radius — dependent analyses / panels / dashboards / alerts / notifications; whether the action is reversible; whether sub-objects are deleted transitively; cross-profile impact.

Read-only commands (get, list, search, path, tree, new-name, *-search, *-query, forecast) do not require this confirmation. For detailed wording and the ready-to-use template, see references/idmp-safety-rules.md.

Two working modes

| Mode | CLI families | Use when |

Reviews

Log in to leave a review.

No reviews yet — be the first.

Explore related

Other things in this space — across every part of the ecosystem, not just skills.