v0.3.0

Added

• speclink context: prints the content of the counterparts linked from a set
  of input files (positional arguments or newline-separated stdin via
  --stdin) — full Markdown sections for doc counterparts, full declarations
  including JSDoc for code counterparts. Default output is Markdown suitable
  for direct injection into an agent prompt; --json emits a machine-readable
  report following schemas/context-output.schema.json. Extraction is
  best-effort: check diagnostics located in the input files are reported on
  stderr (or in the diagnostics field) without affecting the exit code.
• TypeScript scanner: records a declarationRange covering each supported
  declaration including its JSDoc block, backing context content extraction.
• just context: prints the linked counterpart content of the uncommitted
  changes.
• AI integration recipes under docs/integrations/: on-edit counterpart
  awareness and gate triage for Claude Code and Codex, and a CI recipe for
  gating the PR change set and reporting counterpart content.
• Copyable agent hook scripts under examples/hooks/: a PostToolUse hook
  that surfaces linked counterpart content on edit, and a Stop hook that
  reports related --gate findings with the flagged counterparts' content as
  Stop additionalContext.
• Distributable agent skills under templates/skills/: speclink-annotate
  (create @doc/@code link pairs and verify them with speclink check) and
  speclink-sync (triage related --gate findings using speclink context).
  Both are also installed in this repository's .claude/skills/.

Changed

• Markdown section extraction moved from the LSP layer into src/core/ and is
  now shared by LSP hover and speclink context (no behavior change).