SET is the thin orchestration layer in the ABVX stack.
It does not replace coding agents. It decides how repo tooling runs, makes workflow choices explicit, and keeps outputs predictable for CI and downstream agent flows.
agentsgenfor repo docs, pack artifacts, checks, and repo mapsIDfor optional human-context hooks- proof-loop artifacts for durable review state
- registry-driven workflow presets across repos
Use the action in a target repository:
- uses: markoblogo/SET@main
with:
workflow_preset: repo-docs
path: "."If you want a review-first planning pass before touching workflow files:
python3 scripts/plan_config_apply.py markoblogo/<owner>/<repo> --format jsonrepo-docs:init + pack + checksite-ai:repo-docs + site pack + analyze + metaminimal: bootstrap-only baseline
When repo-local ID hooks are enabled, SET now does more than run pre_task.
It captures the resolved human bootstrap order and exports two runtime artifacts into the target repo:
docs/ai/id-bootstrap.jsondocs/ai/id-bootstrap.prompt.md
These packets are built from the resolved ID hook output and are meant for downstream agent runs.
The JSON packet is now treated as a formal runtime interface, documented in docs/id-bootstrap.md, rather than as an action-only implementation detail.
Expected bootstrap order:
soul.mdprofile.core.mdhandshake.md
SET also surfaces that order in the GitHub step summary.
SET owns the orchestration registry and repo-config contract:
- schema:
schema/repo-config.v1.json - docs:
docs/repo-config.md - runtime artifact docs:
docs/id-bootstrap.md - examples:
examples/repo-config.example.json - registry entries:
registry/repos/*.json
Validate the registry with:
python3 scripts/validate_registry.pyscripts/plan_config_apply.py is intentionally review-first:
- shows the proposed workflow as structured output
- exports planning artifacts such as
plan.json,workflow.set.yml,pr-body.md,orchestrator-bundle.json, andproposal-lifecycle.json - can compare against a local repo root for drift
- does not write target repositories directly
orchestrator-bundle.json is the upstream handoff contract for Sortie/Symphony-like runners: it carries repo identity, proposed workflow inputs, repomap policy, optional ID bootstrap hints, proof-loop expectations, recommended review lenses, proposal lifecycle guidance, and one registry-selected capability profile without making SET an agent runner. baseline exports only context/loop review hints; research adds diversity and optional memory guidance; governed-runner adds optional memory plus shadow-first governance; loop-readiness exports a disabled L1/L2 report-first contract. proposal-lifecycle.json spells out the proposal-first flow (run -> retained_output -> inspect -> select/apply/discard). rabbithole.seed.md is also exported as an optional local review seed for human-in-the-loop plan exploration.
The bundle also exports an optional agent_governance_capability: a shadow-first contract for a runner-owned policy decision before significant tool calls, append-only audit records, and per-run tool/cost/latency telemetry. It has no runtime dependency and is disabled by default. See docs/agent-governance-capability-contract.md.
Useful commands:
python3 scripts/plan_config_apply.py markoblogo/lab.abvx
python3 scripts/plan_config_apply.py markoblogo/lab.abvx --dry-run --format json
python3 scripts/plan_config_apply.py markoblogo/lab.abvx --export-dir ./.set-plan
python3 scripts/plan_config_apply.py markoblogo/lab.abvx --repo-root /path/to/labDepending on preset/config, SET can drive production of:
AGENTS.md,RUNBOOK.mdllms.txt/LLMS.mddocs/ai/id-context.jsondocs/ai/id-bootstrap.jsondocs/ai/id-bootstrap.prompt.md- repo maps and AI docs under
docs/ai/ - proof-loop artifacts under
docs/ai/tasks/<task-id>/
agentsgen: repo-scoped agent contextID: portable human contextlab.abvx: public catalog/read-only surface
docs/repo-config.mddocs/id-bootstrap.mddocs/config-apply-planning.mddocs/orchestrator-compatibility.mddocs/context-budget-hint.mddocs/loop-readiness-hint.mddocs/loop-readiness-contract.mddocs/research-diversity-hint.mddocs/references/open-notebook.mddocs/references/agent-orchestrators.mddocs/memory-capability-contract.mddocs/agent-governance-capability-contract.mddocs/capability-profiles.mddocs/llmo-capability-map.mddocs/v0.1-scope.mdCONTRIBUTING.md
