AI Project Orchestrator
Developer Official v1.0.12Multi-agent coordination workflow that reasons over a pasted push summary + open-issues list: structured push review, likely-drift flagging, triage + issue-action planning, and a briefing — via a 3-way review loop (Architect → CTO → Orchestrator)
by skrptiq
Sign in to import this skrpt into your workspace.
Sign inWhat's included 20 nodes
Orchestrator Cycle
Main orchestration workflow: push review → drift check → 3-way review loop → triage → issue management → briefing generation
Architect Review
Produces structured findings with code evidence, severity, classification, and confidence ratings
Briefing Generation
Reads git state, open issues, and cross-repo context to produce a slim BRIEFING.md per repo
CTO Response
Gate step — pressure-tests architect findings: accept, reject, or refine each finding
Issue Management
Determines the GitHub Issue actions to take — comments with commit hashes, closes for single-repo, label changes for cross-repo — for the executing environment to apply
Push Review
Reviews a git push: diffs, commit messages, agent context — produces a structured summary and impact assessment
Schema Drift Check
Reasons over the push summary you provide to flag likely schema/contract divergence that could cause integration failures
Triage Decision
Converts review outcomes into GitHub Issues, watchpoints, or dismissals
Architect Findings
Produces structured review findings with code evidence, severity, classification, and confidence
Check Schema Drift
Compares shared schemas across repos and flags divergence with specific field-level evidence
CTO Pressure Test
Gate prompt — pressure-tests each architect finding and issues a pass/fail verdict for the review loop
Manage Issues
Executes GitHub Issue lifecycle actions: comments, closures, label updates, and new issue creation
Review Push
Analyses a git push and produces a structured summary with impact assessment
Triage Outcomes
Converts accepted review findings into GitHub Issues, watchpoints, or documented dismissals
Briefing Protocol
Rules for context budget, priority ordering, cross-repo fact propagation, and the 140-line insight
Session Protocol
Agent session lifecycle: read briefing, work priority queue, commit, write .orchestrator-msg, push
BRIEFING.md Template
Template for generated briefings — the ~60 lines of session-specific work orders
CLAUDE.md Template
Template for repo-level agent instructions — the ~80 lines an agent reads at every session start
Post-Push Hook Script
Template for the post-push hook that reads .orchestrator-msg and logs to the orchestrator's inbox
Review Record Template
Structured template for the Architect → CTO → Orchestrator review record
Requirements 1 service, 1 permission, 1 data type
- LLM Service
- shell:execute
- source code
Security Passed
Detected
- Services: llm-service
- Permissions: shell:execute
- Data Handling:
Version history 11 releases
GH#858 (A3) — honest copy. The workflow is all-LLM and reasons over a pasted push summary + open-issues list; it never reads a repo or calls the GitHub API. The prior copy over-promised live-repo awareness ("cross-repo drift detection", schema-drift "compares … across repos", issue-management "Updates GitHub Issues" / "Executes"). Tightened the manifest description, the `schema-drift-check` and `issue-management` skill descriptions/capabilities, and the workflow's Overview + issue-management step so it clearly reads as *"reasons over the push summary you provide"* and *"determines the issue actions for the executing environment to apply"* — matching the skill's own existing caveat that it does not call the API. Added an explicit "How it sees your project" note pointing at the `push-hook-script` capture path and at #860 (the planned live-repo MCP / local git-gh capability). Copy only — no workflow, graph, or behaviour change.
GH#844 — migrate the gate step from node-meta (`metadata.gate: true` in the skill) to the canonical execution-entry `gate: true` on the workflow step. Single source of truth; the engine + app read the execution entry. No behaviour change — `IsGate` is identical.
GH#745 — declare per-step `output: {name, type}` on every execution step (push_review/text, drift_check/text, architect_findings/text, cto_response/text, triage/text, issue_actions/text, briefing/text). Lights up the #744 rich flow-map. Content-only; no bindings or logic changes.
GH#645 Row 3b — migrate to K-037 dep-referenced schema. Strip 2 inline shared-content files and declare 2 hub-shared deps (UUID id + slug name + version + checksum from `gen-dep-checksums.mjs`). Internal slug references rewritten for E2 rename/mirror-drop pair(s): generate-briefing→agent-session-briefing. Closes pre-Step-3 inline-vendoring for this bundle.
Wave 2: re-signed with canonical engine signing pipeline.
Signature fix — RELEASE_NOTES.md now included in integrity checksum.
Initial catalogue release with full structural and content-quality validation. All scanner checks pass.
Release notes unavailable for this version.
Release notes unavailable for this version.
Release notes unavailable for this version.
Release notes unavailable for this version.
More from Developer
View all →Release Process Pipeline
Drafts user-facing release notes from recent commits and pull requests
Documentation Pipeline
Analyzes source code and generates comprehensive technical documentation with examples, formatted as clean markdown
Skrpt Builder
Design and generate a complete skrpt from a plain-language description — the guided way to create your first skrpt