Orchestration

GHS provides two orchestration skills for automated, end-to-end workflows that chain individual skills together.

Multi-Repo Pipeline (ghs-orchestrate)

Chain skills into a sequential pipeline across one or many repositories:

When to Use

• You manage multiple repositories and want to maintain them all in one session
• You want a full end-to-end cycle: scan, fix, review, merge, sync, and release
• You need resume-from-interruption support for long-running maintenance

Key Features

• 7-stage pipeline: pull, scan, fix, review, merge, sync, release
• Human checkpoints: Pause for confirmation before destructive stages (fix, merge, release)
• State issue resume: Pick up where you left off if interrupted (session state stored in a GitHub Issue with ghs:state label)
• Dry-run mode: Preview what would happen without executing
• Partial ranges: Run only specific stages with --from / --to

Example Session

You: maintain my repos
GHS: [shows pipeline plan for 3 repos, asks confirmation]

You: y
GHS: [pulls all repos]
GHS: [scans repo 1: 8 FAIL items]
GHS: ## Checkpoint: Fix Stage --- repo 1
     8 items to fix. This will create PRs. Proceed? (y/n/skip)

You: y
GHS: [fixes 5 items, creates 3 PRs]
GHS: [reviews 3 PRs]
GHS: ## Checkpoint: Merge Stage --- repo 1
     3 PRs approved. Proceed? (y/n/skip)

You: y
GHS: [merges 3 PRs, syncs remaining items]
GHS: [moves to repo 2...]

Single-Repo Developer (ghs-dev-loop)

Act as an autonomous developer for one repository, processing work in priority-driven cycles:

When to Use

• You want GHS to handle everything for a single repository
• You have a backlog of issues and want them triaged, analyzed, implemented, and merged
• You want priority-based processing: critical bugs before features, health before issues

Key Features

• Priority queue: P0 (critical health) through P6 (release), processed in order
• 3 cycle modes: single (one pass), continuous (until empty), watch (poll for new issues)
• Budget control: Limit issues processed per cycle (default: 5)
• 4 checkpoint gates: health fix, implement (High+), merge, release
• Circuit breaker: 3 failed implementations stops the cycle

Example Session

You: dev loop on phmatray/Formidable
GHS: [pre-flight checks, loads state issue]
GHS: ## Priority Queue --- phmatray/Formidable
     Health Score: 65% (P2 maintenance needed)
     Open Issues: 7 (1 critical, 2 high, 4 normal)
     Bot PRs: 3 pending
     Mode: single (budget: 5 issues/cycle)
     Proceed? (y/n/adjust)

You: y
GHS: [fixes health: 65% -> 82%]
GHS: [processes critical issue #12: analyze -> implement -> review -> merge]
GHS: [processes high issue #15: analyze -> implement -> review -> merge]
GHS: ## Checkpoint: Implement #18
     Complexity: VeryHigh. Proceed? (y/skip)

You: skip
GHS: [skips #18, processes normal issue #22]
GHS: [merges 3 Renovate PRs]
GHS: ## Cycle Summary
     Health: 65% -> 82% (+17%)
     Issues: 4 processed, 3 merged, 1 deferred
     PRs: 3 created, 8 merged

Comparison

Feature	ghs-orchestrate	ghs-dev-loop
Scope	Multi-repo	Single-repo
Focus	Health pipeline	Full issue lifecycle
Issue processing	No	Yes (triage, analyze, implement)
Code review	Yes	Yes
Priority queue	No (sequential stages)	Yes (P0-P6)
Resume	State issue per repo	State issue per cycle
Modes	Single run	Single, continuous, watch
Budget	N/A	Configurable per cycle