Releases: ghspain/github-build-or-reuse
Release list
GitHub Build or Reuse v1.2.0
v1.2.0 — Search before you build
This release sharpens GitHub Build or Reuse as a pre-build gate for AI coding agents: broader activation, clearer due diligence behavior, substantially improved discoverability, and a real reproducible demo instead of an illustrative one.
Highlights
- Broader automatic activation. The skill now explicitly catches common solved-problem categories such as authentication, payments, scraping, browser automation, notifications, search/RAG, observability, GitHub automation, media processing, queues, schedulers, rate limiting and workflow engines.
- No outsourced due diligence. When tools can verify license, maintenance, architecture or security evidence, the skill performs that work instead of handing it back to the user.
- No silent degradation. Missing structured checks are named and confidence is lowered rather than presenting weaker fallback research as equivalent evidence.
- Better GitHub/open-source positioning. The README and plugin metadata make open-source alternatives, software reuse and build-vs-open-source decisions easier to discover.
- Real evidence demo. The README scenario uses a dated public-repository snapshot for an AI presentation-generator decision and derives its SVG, Asciinema recording and animated GIF from one auditable JSON source.
- Reproducible visuals. The demo generator uses only the Python standard library; CI detects drift, while the animated GIF reuses the official
asciinema/aggrenderer pinned by version and checksum.
The decision model remains stable
The core contract is unchanged:
USEan existing project when it clears the important requirements and risk gates.CONTRIBUTEwhen a strong upstream exists and the missing work belongs there.FORKwhen a strong base exists but sustained divergence is intentional.BUILDwhen no candidate clears the gates or greenfield ownership is genuinely the better option.
This remains deliberately broader than a simple Free / Build / Buy comparison because upstream contribution and deliberate fork ownership are first-class engineering strategies.
Install
Latest version:
gh skill install ghspain/github-build-or-reuse github-build-or-reusePin this release:
gh skill install ghspain/github-build-or-reuse github-build-or-reuse@v1.2.0Using the cross-agent skills CLI:
npx skills@latest add ghspain/github-build-or-reuse --skill github-build-or-reuseFor the Codex plugin marketplace:
codex plugin marketplace add ghspain/github-build-or-reuse --ref v1.2.0Demo evidence
The included README demo is a dated example, not a benchmark or permanent project endorsement. Its evidence source lives in assets/demo.json; regenerate the deterministic assets with:
python scripts/generate_demo.py
python scripts/generate_demo.py --checkRelated work
The README acknowledges both polmarza/github-repo-scout and Emanuelel/dont-reinvent. The implementation, decision model and visual assets remain independent.
GitHub Build or Reuse v1.1.0
v1.1.0 — Portable skill + plugin packaging
This release turns GitHub Build or Reuse into a portable Agent Skills project with OpenAI plugin packaging while keeping one canonical runtime skill.
Highlights
- Canonical Agent Skill under
skills/github-build-or-reuse/. - OpenAI ChatGPT/Codex plugin manifest and Git-backed marketplace entry.
- GitHub CLI /
gh skillinstallation and update path. - Output evals plus trigger and non-trigger queries.
- Stronger handling of license, maintenance, security, governance and adoption cost.
- Explicit protection against treating candidate repository instructions as trusted authority.
- GitHub Community Spain as the authoritative project publisher and lifecycle owner.
Install
Latest compatible version:
gh skill install ghspain/github-build-or-reuse github-build-or-reusePin this release:
gh skill install ghspain/github-build-or-reuse github-build-or-reuse@v1.1.0For ChatGPT/Codex plugin installation, use the Git-backed marketplace documented in the README.
Architecture decision
No custom GitHub MCP server is bundled. Existing GitHub connectors/API, gh, MCP integrations, or web research already provide the required evidence surface. A bespoke MCP will only be added when it closes a demonstrated cross-host capability gap.