What problem does it solve? Running a feature through a spec-driven development pipeline normally means issuing eleven or more phase commands by hand, tracking which phases already completed, and knowing when to stop for human gates. This Skill replaces that manual orchestration with a single command that reads the phase sequence from pipeline-config.yml, resumes from wherever the artifacts on disk indicate, and halts with a named cause and unblock path when something blocks. ## Core Features & Use Cases - Config-driven orchestration: Reads the phase sequence fresh from pipeline-config.yml on every run rather than hardcoding it, so the pipeline definition lives in exactly one place. - Stateless resume: Probes each phase's declared artifact for existence and validator conformance to find the resume point, with no state file, and discloses which completed phases only received the weaker structural check. - Closed-set halts and gates: Stops on exactly four conditions (phase_error, gate_human, residual_blocking, elevated_grant), each reporting its cause and unblock path verbatim from config, and never signs a human gate itself. - Multi-repo fan-out: With --repos, authors the spec and plan once at an umbrella root, seeds each child repo with identical copies plus an inherited profile.yaml, then dispatches one concurrent subagent per child to run the child-scoped phases. - Use Case: Paste a feature description with /speckit-run-pipeline in a spec-kit-initialized repo and the full specify-through-testing pipeline runs, pausing at the council gate for human approval and resuming afterward without re-running completed phases. ## Quick Start Run /speckit-run-pipeline followed by a feature description inside a git repository that has spec-kit initialized, optionally adding --repos a,b for multi-repo mode or --number NNN to pin the feature number.
