abvx-set 0.2.4

SET workflow action helpers and registry tooling

SET

Thin orchestration repo for the ABVX stack to keep AI-facing repo artifacts in sync.

SET is not a replacement for Copilot/Cursor/Continue. It is a workflow control plane: it orchestrates existing tools, makes run plans explicit, and keeps outputs predictable for CI and agents.

Quick start (30 seconds)

Use this in any repo to baseline AGENTS.md and checks via agentsgen:

- uses: markoblogo/SET@v0.1.0
  with:
    workflow_preset: repo-docs
    path: "."

If you're not ready to wire .github/workflows/set.yml manually, run plan-config locally first:

python3 scripts/plan_config_apply.py markoblogo/<owner/repo> --format json

This is a read-only plan showing the exact workflow file that would be generated.

Who this is for

• Solo devs: one-command AI-ready repo baseline (agentsgen, checks, repomap, optional proofs).
• Team maintainers: one central registry for many repos with CI-drift visibility.
• Tool builders: stable registry and planner artifacts for control-plane integrations.

What SET Does

• agentsgen handles repo documentation, map generation, checks, and related contracts.
• SET chooses how and when agentsgen and other tools run.
• ID and other optional integrations are executed via explicit, validated repo config.
• lab.abvx is the public catalog and read-only control-plane surface.

flowchart LR
  SET["SET Action"]
  agentsgen["agentsgen"] -->|generates docs| artifacts["AGENTS.md / llms.txt / docs/ai"]
  SET --> agentsgen
  ID["ID"]
  ID -->|hooks / context| SET
  SET --> lab["lab.abvx"]
  registry["Registry"] --> SET

Core roles

• repo-docs preset: init + pack + check
• site-ai preset: repo-docs + site pack + analyze + meta
• minimal preset: bootstrap only

Repo config contract

SET owns the first centralized contract and registry baseline for repo-level orchestration.

• Contract docs: docs/repo-config.md
• Schema: schema/repo-config.v1.json
• Registry: registry/repos/*.json
• Validation: python3 scripts/validate_registry.py

The contract includes:

• tooling intent (tools.agentsgen, tools.id, tools.git_tweet)
• presets and overrides (workflow_preset inputs)
• repomap policy (compact_budget, top_ranked_files, focus, changed)
• proof-loop contract metadata
• optional site baseline for catalog-facing pages

There is an additional runtime artifact, agents.knowledge.json, that is generated by agentsgen and consumed by AI tooling. It is intentionally separate from control-plane registry data.

Config apply planning (scripts/plan_config_apply.py)

Current behavior is intentionally planning-only:

• shows one-file proposed workflow diff as structured output
• exports review artifacts (plan.json, workflow.set.yml, pr-body.md, etc.)
• performs repo drift check with --repo-root (matches, drift, missing)
• never writes target repos (review-first only)

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/lab

Getting started docs

• docs/repo-config.md
• docs/config-apply-planning.md
• docs/llmo-capability-map.md
• docs/v0.1-scope.md
• CONTRIBUTING.md

Links

• SET repo: https://github.com/markoblogo/SET
• Lab catalog: https://github.com/markoblogo/lab.abvx
• ID protocol: https://github.com/markoblogo/ID
• agentsgen: https://github.com/markoblogo/AGENTS.md_generator