{"id":43228726,"url":"https://github.com/tim-osterhus/olad-framework","last_synced_at":"2026-02-07T06:03:15.968Z","repository":{"id":334446171,"uuid":"1137240781","full_name":"tim-osterhus/olad-framework","owner":"tim-osterhus","description":"Orchestrate agentic development for hours with zero scripting: markdown-driven roles, fresh context per task, \"blind\" QA validation, and on-demand automatic skill generation—all while remaining vendor-agnostic, deterministic, and auditable.","archived":false,"fork":false,"pushed_at":"2026-01-25T04:08:45.000Z","size":127,"stargazers_count":2,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-01-25T08:40:23.843Z","etag":null,"topics":["agentic-ai","agentic-coding","agentic-development","agentic-framework","agentic-workflow","customizable","markdown-language","vendor-agnostic"],"latest_commit_sha":null,"homepage":"","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/tim-osterhus.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-01-19T05:34:40.000Z","updated_at":"2026-01-25T04:10:11.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/tim-osterhus/olad-framework","commit_stats":null,"previous_names":["tim-osterhus/olad-framework"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/tim-osterhus/olad-framework","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tim-osterhus%2Folad-framework","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tim-osterhus%2Folad-framework/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tim-osterhus%2Folad-framework/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tim-osterhus%2Folad-framework/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/tim-osterhus","download_url":"https://codeload.github.com/tim-osterhus/olad-framework/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tim-osterhus%2Folad-framework/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":28975278,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-01T09:57:52.632Z","status":"ssl_error","status_checked_at":"2026-02-01T09:57:49.143Z","response_time":56,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agentic-ai","agentic-coding","agentic-development","agentic-framework","agentic-workflow","customizable","markdown-language","vendor-agnostic"],"created_at":"2026-02-01T10:01:16.570Z","updated_at":"2026-02-01T10:01:17.339Z","avatar_url":"https://github.com/tim-osterhus.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"![License](https://img.shields.io/badge/license-Apache%202.0-blue)\n![Status](https://img.shields.io/badge/status-active-success)\n![Contributions by invite](https://img.shields.io/badge/Contributions-invite%20only-orange)\n\n# OLAD Framework\n\nOrchestrative Lean Agentic Development (OLAD) is a lightweight, project-agnostic framework for running agentic development with clear roles, repeatable prompts, deterministic orchestration, and consistent logging.\n\nTo get started ASAP, check out `quickstart.md`.\n\n## What This Gives You\n\n- A standard way to run multi-agent work without chaos.\n- A stable structure for task prompts, QA, and history logging.\n- A deterministic orchestration loop with clear status signaling.\n- A repeatable onboarding flow that makes a repo “agent-ready.”\n\n## How It Works (High Level)\n\nThe framework defines a small set of agent entrypoints and artifacts:\n\n- **Builder**: implements tasks and small fixes quickly and safely.\n- **QA**: verifies changes and records evidence.\n- **Advisor**: handles freeform tasks like scoping work, explaining code, or drafting backlog task cards.\n- **Orchestrator**: runs Builder → QA → (optional) Quickfix headlessly. Optional orchestration features (e.g., Integration cycles, Troubleshooter, no-manual QA smoketests) are enabled during customization.\n\nBuilder runs create and consume prompt artifacts by default, and every outcome is recorded at the top of `agents/historylog.md` (newest first). Orchestration uses `agents/status.md` as the sole signaling file.\n\nPer-cycle runners (Codex CLI / Claude Code / OpenClaw) and model choices are configured in `agents/options/model_config.md`.\n\nOperational spec: see `OLAD_framework.md`.\n\n## Why It’s Useful\n\n- **Clarity**: every agent knows where to look and how to proceed.\n- **Traceability**: work is logged with context and evidence.\n- **Repeatability**: you can reuse the workflow across repos and teams.\n- **Determinism**: orchestration is driven by explicit status flags, not implicit heuristics.\n\n## One-Time Setup (Recommended)\n\nUse this once per repo to make the framework project-specific.\n\n1) Start an agentic coding session at the root of your target repo.\n2) Say: `Open agents/_customize.md and follow instructions.`\n3) The agent generates `agents/spec.md` and fills out key project files.\n\nDuring customization, the agent will also ask you which **model preset** you want and update `agents/options/model_config.md` accordingly. It will also set headless sub-agent permissions (Normal/Elevated/Maximum) in `agents/options/workflow_config.md`. Orchestrator behavior is configured during customization via `agents/options/`.\n\nOptional feature packets and templates live under `agents/options/` to keep entrypoints small.\n\nAfter setup, the repo is ready for day-to-day agentic work.\n\n## Project Context (fill during customization)\n\n- **What this repo is:** \u003cone sentence describing the product/system and its purpose\u003e.\n- **Production target:** \u003cdeployment constraints, e.g., offline-only, regulated, latency goals\u003e.\n- **Guardrails:** \u003cdata handling, safety/compliance, or review requirements\u003e.\n- **Review gates:** \u003chuman review or QA gates required before shipping\u003e.\n\n## Non-negotiables (safety + quality)\n\n- **Environment constraints:** respect deployment limits (offline, air-gap, latency, cost).\n- **Data handling:** follow privacy, security, and compliance requirements.\n- **Quality gates:** do not ship without required verification and review.\n- **No secrets** in repo or logs.\n\n## Detailed Features and Workflows (In Full-Cycle Order)\n\nThis section follows the sequence you would encounter in a real workflow cycle.\n\n### 1) Scope the Work (Advisor Session)\n\nThe Advisor is a dedicated, freeform session used to make the rest of the cycle clean and scoped.\n\nIf you want a more structured, high-output scoping pass (raw idea -\u003e spec -\u003e ordered task cards), use:\n- `agents/prompts/decompose.md` (via the Advisor)\n\nTypical Advisor outputs:\n- A set of task cards that are small and executable.\n- A short explanation of how a subsystem works.\n- A recommendation memo with options and tradeoffs.\n\nIf you want a more structured \"idea -\u003e spec -\u003e ordered backlog\" flow, use `agents/prompts/decompose.md` instead.\n\nThe Advisor should use `agents/_advisor.md`, rely on `agents/outline.md` as the primary repo overview, and place any task cards in `agents/tasksbacklog.md` (not `agents/prompts/tasks/`).\nAdvisor runs only log to `agents/historylog.md` when they perform concrete actions beyond writing task cards (prepend new entries at the top).\nWhen needed, the Advisor can also generate new roles and skills using the dedicated creation prompts.\n\n### 2) Orchestrate the Cycle (Orchestrator Session)\n\nThe Orchestrator runs Builder → QA → (optional) Quickfix automatically against the backlog. Optional steps (like Integration cycles or Troubleshooter) are installed during customization.\n\nKey behavior:\n- Pulls the next task from `agents/tasksbacklog.md` into `agents/tasks.md`.\n- Runs Builder and QA in strict order, with optional steps installed by customization when enabled.\n- Allows a Quickfix pass if QA finds gaps.\n- Archives completed tasks to `agents/tasksarchive.md`.\n- Clears `agents/status.md` to `### IDLE` after each action.\n\nUse `agents/_orchestrate.md` for the exact runbook and guardrails.\n\n### 3) Prompt Artifacts (Prompt Engineering Cycle)\n\nPrompt artifacts are the default plan carrier for each task. They can be executed without re-planning.\n\nCore flow:\n1) Create a task prompt with `agents/prompts/create_prompt.md`.\n2) Store it in `agents/prompts/tasks/`.\n3) Execute it with `agents/prompts/run_prompt.md`.\n4) Archive it to `agents/prompts/completed/` after execution.\n\nThis makes a prompt-driven workflow repeatable and reviewable.\n\n### 4) Build and Quickfix (Builder Session)\n\nThe Builder executes the task and makes the smallest viable change set. It always creates a prompt artifact first and uses it as the authoritative plan.\n\nBuilder guidance:\n- Use `agents/_start.md`.\n- Keep changes minimal and scoped to the task.\n- Prefer safe, reviewable diffs.\n\nQuickfixes use `agents/_hotfix.md` and follow `agents/prompts/quickfix.md` for narrow, urgent changes.\n\n### 5) QA and Verification (QA Session)\n\nQA is a first-class session with its own entrypoint and verification flow.\n\nKey pieces:\n- `agents/_check.md` defines the QA flow.\n- `agents/expectations.md` is an explicit expectations list used to anchor QA. This is a deliberate feature: it produces measurably stronger QA output because the agent verifies against concrete, written expectations rather than generic “check for issues.”\n- QA results should be logged with evidence (commands, outputs, or screenshots).\n\nDoublecheck QA uses `agents/_doublecheck.md` to validate quickfixes.\n\n### 5.5) Integration (Optional)\n\n- **Integration Steward:** `agents/_integrate.md` runs an integration sweep and writes an Integration Report. Orchestrated Integration runs only if enabled during customization, and tasks may use `**Gates:** INTEGRATION` only when Integration is installed.\n\n### 5.6) No-Manual QA (Optional)\n\nIf enabled during customization, QA must not request human/manual verification and instead replaces manual checks with tracked smoketest artifacts under `agents/prompts/tests/`.\n\n### 6) Roles and Specialization\n\nRoles are defined in `agents/roles/` and used when a task benefits from a specific lens (security, infra, QA, etc.). Use one role at a time, and switch explicitly.\n\nRoles make multi-agent collaboration more predictable and reduce churn.\nIf a new role is needed, use `agents/prompts/role_create.md` to generate a role file that matches the repo’s conventions.\nAlternatively, use `agents/prompts/roleplay.md` for an interactive, ultra-detailed role generator.\nMost roles are broadly reusable across projects.\n\n### 7) Skills (Reusable Procedures)\n\nSkills are reusable playbooks stored under `agents/skills/`.\n\nHow they work:\n- Each skill defines a specific workflow, constraints, and outputs.\n- If a task matches a skill, the agent should follow it rather than improvising.\n- Skills are meant to reduce risk and improve consistency across repos.\nIf a new skill is needed, use `agents/prompts/skill_issue.md` for the full, interactive skill generator.\nNote: several starter skills are often project-specific and can be removed during customization.\n\n### 8) History Logging\n\nEvery session prepends to the top of `agents/historylog.md` (newest first) with a short summary, files touched, decisions, and follow-ups. This makes agent work auditable and easy to resume.\n\n## Files Worth Reading\n\n- `quickstart.md` — step-by-step usage guide\n- `agents/_customize.md` — one-time onboarding prompt\n- `agents/_orchestrate.md` — orchestration entrypoint\n- `agents/options/` — optional feature packets, configs, and headless templates\n- `OLAD_framework.md` — detailed orchestration model and contracts\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftim-osterhus%2Folad-framework","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftim-osterhus%2Folad-framework","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftim-osterhus%2Folad-framework/lists"}