Task2Tool v0.3.0 — Coverage-aware capability composition

Compose a compact set of complementary local skills, prompts, agents, MCP servers, and catalogs for a task.

Highlights:

• Greedy marginal lexical coverage with deterministic tie-breaking.
• Explains newly covered terms, cumulative coverage, and explicit remaining gaps.
• Schema 1.2 discloses bounded query evaluation so truncated input can never appear as complete coverage.

Proof boundary: coverage is normalized lexical overlap, not semantic planning or a capability guarantee.

What's Changed

• feat(scanner): discover GitHub Copilot instruction Markdown files by @dchaudhari7177 in #3
• scanner: parse YAML block-list tags/keywords in frontmatter by @dchaudhari7177 in #5

New Contributors

• @dchaudhari7177 made their first contribution in #3

Full Changelog: v0.2.0...v0.3.0

Task2Tool v0.2.0 — Instant local AI capability discovery

Task2Tool v0.2.0 makes local capability discovery useful from the first command.

Highlights

• Added task2tool demo with bundled skills, prompts, agents, MCP configuration, and catalog examples.
• Added an optional natural-language demo query with ranked, explainable results.
• Reworked the quick start and contribution paths around a zero-setup first run.
• Added branded README and social-preview artwork.

Try it:

npx --yes github:mockingbird777/task2tool#v0.2.0 demo

Task2Tool remains local-first, has zero runtime dependencies, and requires Node.js 20 or newer.

Task2Tool v0.1.0

First public release of Task2Tool.\n\nHighlights:\n- Deterministic task-to-capability search across skills, prompts, agents, MCP configurations, and portable catalogs\n- Local-only index, find, and lint workflows with zero runtime dependencies\n- JSON, Markdown, and self-contained HTML reports\n- CJK-aware tokenization, bounded scanning, secret-safe MCP metadata, and atomic output\n- Interactive Pages demo and CI-tested GitHub installation path\n\nRequires Node.js 20 or newer.