Agentic Project Lifecycle is a cross-platform skill suite for governing large software projects from discovery through production operations. It combines one lifecycle orchestrator with six focused specialist skills, executable project fixtures, held-out routing tests, and a measurable release-readiness gate. Codex is one supported native distribution target, not the only supported platform.
Release status:
v1.1.0-rc.1is a published prerelease with deterministic assets, SHA-256 checksums, and provenance/SBOM attestations. Stable1.1.0remains blocked on authenticated repeated live-agent evaluation.
English | Русский | Українська
- Turn a product idea into a governed delivery plan with explicit artifacts and decision gates.
- Build SaaS and AI products with domain-specific product, architecture, evaluation, and operational controls.
- Modernize brownfield systems while preserving compatibility and planning safe cutovers.
- Rescue delayed or unstable projects using evidence-first triage and rebaselining.
- Prepare releases, rollback plans, observability, and incident-learning loops.
- Audit traceability, verification evidence, and release readiness without inventing missing proof.
| Skill | Purpose |
|---|---|
orchestrating-large-projects |
Routes work across the lifecycle and keeps project artifacts coherent. |
building-saas-products |
Covers SaaS discovery, tenancy, entitlements, billing, activation, and operations. |
building-ai-products |
Defines AI product boundaries, evaluations, safety controls, and operations. |
modernizing-existing-projects |
Handles brownfield discovery, migration patterns, cutover, and decommissioning. |
rescuing-software-projects |
Triages distressed projects and creates evidence-based recovery plans. |
releasing-and-operating-products |
Governs staged releases, reliability, rollback, incidents, and learning. |
auditing-project-readiness |
Validates project state, traceability, evidence, and release readiness. |
Add this repository as a plugin marketplace, then install the plugin:
codex plugin marketplace add VKK-00/agentic-project-lifecycle --ref v1.1.0-rc.1
codex plugin add agentic-project-lifecycle@vkk-00-agent-plugins
In the Codex desktop app, the same marketplace can be added from Plugins > Marketplaces, after which Agentic Project Lifecycle appears in the plugin list.
APL has one canonical seven-skill inventory and produces deterministic bundles for Agent Skills, Codex, Claude Code, GitHub Copilot, Cursor, Kimi Code, Gemini CLI, OpenCode, Factory Droid, Amp, Devin, Pi, Hermes, Antigravity, and Gemini Enterprise.
To inspect supported targets and install into an explicit project or user root:
python apl platform list
python apl platform install codex --scope project --root /path/to/project
python apl platform verify /path/to/project/.codex/skills/agentic-project-lifecycleThe installer stages and verifies a full copy before atomically publishing it; --force restores the previous installation if post-publication verification fails. --scope user always requires an explicit --root; it never writes to a guessed home directory.
Ask Codex something concrete, for example:
Turn this product idea into a governed project from discovery through release.
Audit this project's readiness for beta or general availability.
Create a recovery plan for this delayed software project.
The orchestrator selects the relevant specialist skills. You can also invoke a skill explicitly, for example $auditing-project-readiness.
Requirements: Python 3.11 or newer.
python -m pip install -e ".[dev]"
python validate.py
python -m pytest -q
python scripts/validate_publication.pyBuild deterministic release archives:
python scripts/build_release.py --version 1.1.0-rc.1 --output distThe build creates ZIP and tar.gz plugin archives, 15 platform bundles, validation evidence, the promotion-gate result, and SHA-256 checksums.
Distribution support does not prove that a platform activated the skills or that a model followed them. Activation evidence and bounded execution are separate claims; the committed activation matrix intentionally records every platform as not-live-tested.
The repository includes held-out routing cases, executable fixture projects, pinned read-only public-repository trials, execution-trace analysis, and leave-one-rule-out instruction ablation. See Validation, Stability report, and evaluation documentation.
These evaluations test the suite's routing and artifact rules; they do not prove that every model run or project outcome will be correct. The public-repository trials do not certify the sampled projects. Users remain responsible for reviewing changes and approving consequential actions.
Licensed under Apache License 2.0. See LICENSE and NOTICE.
This is an independent community project maintained by VKK-00. It is not an official OpenAI product. The project was developed with assistance from OpenAI Codex; responsibility for the published content remains with the maintainer.