A cross-agent upgrade tool promising a reviewable diff and no lost local edits.
A portable bundle carries metadata and relative UTF-8 files. Treat SKILL.md as an entry document and reject any resource name that could escape the selected install root. The original Orchard bundle includes a restore checklist.
Start with a learner workspace
Data management, Security and secrets audit. Language foundation: TypeScript object types.
Use Node 22.18+ and Python 3.10+ for the grader. Core TypeScript runs without package installation.
python3 scripts/project_test.py skill-installer --init learning-artifacts/skill-installer
python3 scripts/project_test.py skill-installer --stage 1 --path learning-artifacts/skill-installerBuild route
- Validate a portable bundle
- Translate metadata and hash content
- Install atomically within a root
- Protect edits during upgrades
Run with your own inputs
After completing the stages, these commands run your workspace code on the original Orchard examples. Replace the sample paths with your own files.
node learning-artifacts/skill-installer/cli.ts inspect projects/skill-installer/examples/orchard-release.json codexTo inspect the complete reference first, replace learning-artifacts/skill-installer with projects/skill-installer/solution in the same command. JSON results use schema_version: 1; paths and argument examples are explicit so another tool can consume them.
Integration boundary
inspect is a dry run. install needs bundle, agent, root and trusted expected source digest. Supported destination mappings are codex=.agents/skills, claude=.claude/skills and cursor=.cursor/skills. Discovery paths are configured locally; successful installation does not prove a separately running agent activated the skill.
python3 scripts/project_test.py skill-installer --all --solution --strict
python3 scripts/project_test.py skill-installer --all --path learning-artifacts/skill-installer --strictThe first command checks the reference. The second checks your implementation. The published examples and tests are regression evidence, not a production certification or an unseen benchmark.
