← ClaudeAtlas

pending-approvalslisted

Use when the user invokes approvals directly -- "ver pendientes", "aprobar P-XXXX", "rechazar P-XXXX", "approve P-", "reject P-"
metraton/gaia · ★ 3 · AI & Automation · score 76
Install: claude install-skill metraton/gaia
# Pending Approvals `pending-approvals` is the workflow the orchestrator follows when the *user* drives an approval -- "ver pendientes", "aprobar P-XXXX", "rechazar P-XXXX". Approvals are **in-loop and single-session**: a pending is raised and resolved within the same session, and there is no proactive surfacing of it outside that flow -- no `[ACTIONABLE]` block at SessionStart, no per-turn resurfacing. The orchestrator's role here is to translate an explicit user ask into the right `gaia approvals` subcommand against the right store, then get it run. See "Approvals do not resurface across sessions" below for what does and does not survive cross-session. **Execution: dispatch, don't run directly.** The orchestrator carries no shell. Every `gaia approvals` subcommand named in this skill (`show`, `approve`, `revoke`, `reject`, `reject-all`, `clean`, `pending`) is executed by dispatching a subagent -- gaia-operator by default, the general-purpose executor -- with the exact command; the subagent returns the output for the orchestrator to relay, verify, or build the `AskUserQuestion` payload from. Wherever this skill says "run" or "call" a `gaia approvals` subcommand, read it as "dispatch gaia-operator to run" -- the orchestrator's job is choosing the right subcommand and store and acting on what comes back, never executing it itself. For the universal envelope of an approval payload see `agent-approval-protocol`; for how the orchestrator relays a *subagent-initiated* APPROVAL_