Skip to content
Agentmart
Submit a tool
This entry duplicates portable-application, which is the one we list in search.

portable-application

Portable complete application generation.

Skill by Literate AI maintainers <literate-ai-maintainers@users.noreply.github.com>

IndexedBSD-2-Clause
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/jordanhubbard-portable-application-d2e6ff

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Portable complete application generation. Use for Literate AI workflow tasks.

Use cases: productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions: not checked yet.

What's inside

Skill name
portable-application
Description
Portable complete application generation. Use for Literate AI workflow tasks.
License
BSD-2-Clause
Metadata
author: Literate AI maintainers <literate-ai-maintainers@users.noreply.github.com>
Instructions
10 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Portable complete application generation

Generate a complete application from the exact accepted specifications and selected Flavors. When a literate specification-context document is present, preserve its node, ancestor, and reference provenance, read each original document once, and surface contradictions rather than silently erasing a normative statement through nearest-wins precedence. Keep this Component's reasoning context bounded to local authority and direct dependencies' public interfaces; never import private or transitive implementation material. Preserve every normative observable behavior, emit all implementation source and current tests, and provide a runnable entrypoint named run. The entrypoint must accept one strict JSON value as its first command-line argument and emit exactly one strict JSON value on standard output.

The runnable artifact SHALL take exactly one command-line argument: a complete UTF-8 JSON **array** of the application's arguments, and SHALL spread that array's elements as the application's positional arguments. A Component whose specification describes "one JSON object as its first argument" therefore receives `[{...}]` on the command line and passes the single object through, rather than receiving the bare object. Emit exactly one JSON result plus a trailing newline on standard output, and send diagnostics only to standard error. This is the same contract the independent acceptance verifier uses to invoke the built artifact, so an implementation that accepts the bare object instead of the array will pass its own generated tests and still fail acceptance.

Before product argument parsing, support `--litai-test` by running every current generated native case. Write `source/tests/manifest.json` as compact canonical JSON with object keys sorted lexicographically and no insignificant whitespace. On complete success this mode must emit exactly one compact JSON object of the form `{"schema":"literate-ai/generated-test-results@1","cases":[{"case_id":"CASE-ID","outcome":"passed"}]}`. Include one unique case entry for every and only case in the manifest, preserve its case IDs exactly, use the literal outcome `passed`, and emit no other standard output. A failed or unexecuted case must make the mode exit nonzero and must never be reported as passed. Do not substitute newline-delimited records, a `passed` Boolean, test-framework prose, or the manifest itself for this protocol.

Support `--litai-smoke` by exercising one current non-acceptance generated example through real application logic and emitting that example's ordinary JSON result on standard output. This mode must produce observable standard output: the Standard lifecycle executes the built artifact in this mode after source custody is gone and rejects a run that emits nothing, so a silent smoke mode fails the lifecycle even when the application is correct. Neither mode is product behavior and neither may read a generated manifest at runtime. Keep diagnostics on standard error. When the Bazel Flavor is selected, expose the application as //:run, the selected command profile's self-contained export as //:litai_artifact, and tests through bazel test //.... Apply reusable source-disposal, test, portability, and build guidance from this skill and selected Flavors without inventing duplicate product requirements. Agent-authored glue may connect selected interfaces but new observable behavior requires a proposed specification change. Do not consult or reproduce an original source checkout: only specifications, Flavors, and exact skills are implementation inputs.

From jordanhubbard/project-tracker, BSD-2-Clause.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Sep 13, 2026 (1 month ago)
Last release
Sep 20, 2026 (1 month ago)

From GitHub (fetched Oct 24, 2026).

Details

License
BSD-2-Clause
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

Same skill elsewhere

Copies of this SKILL.md published by others; search shows only one of them.

More from Literate AI maintainers <literate-ai-maintainers@users.noreply.github.com>

Cite this page

portable-application skill. Agent Mart. https://agentmart.dev/skills/jordanhubbard-portable-application-d2e6ff (last updated 2026-10-24)