Skip to content

Releases: word-ful/editorial

v0.3.0 — Plugin marketplace install

Choose a tag to compare

@word-ful word-ful released this 17 May 13:24

Two-command install, no terminal required. The skill is now a proper Claude Code plugin.

Added

  • Plugin marketplace install path. Install Dissolve AI from inside Claude Code with two commands:

    /plugin marketplace add word-ful/editorial
    /plugin install dissolve-ai@editorial
    

    No git, no symlinks, no restart. Auto-updates apply on future versions.

Changed

  • disable-model-invocation: true added to dissolve-ai/SKILL.md frontmatter. The skill no longer registers for Claude's auto-trigger consideration — it runs only when the user explicitly invokes it. This was the design intent from v0.2; the explicit flag enforces it.
  • Both READMEs now lead with the plugin install. Manual install (clone + symlink) is still supported as the "Alternative" path.

Bumped

  • Roadmap: v0.4 is now bring-your-own-sample mode.

See CHANGELOG.md.

v0.2.2 — Platform-agnostic positioning + correctness disclaimer

Choose a tag to compare

@word-ful word-ful released this 17 May 12:36

Two small clarifications.

Changed

  • Top-line positioning — dropped "Claude Code" from the READMEs' description sentences. The skill is now described as "a skill" rather than "a Claude Code skill," keeping the brand framing platform-agnostic. Technical install/usage instructions still reference Claude Code where accurate.
  • Correctness disclaimer — added a bullet to the What this skill does NOT do section: It does not rewrite your text to sound "correct." Clarifies that Dissolve AI removes patterns; it does not impose a house style or grade prose toward any standard of correctness.

See CHANGELOG.md.

v0.2.1 — Structure-protection fix + new example

Choose a tag to compare

@word-ful word-ful released this 17 May 11:30

Closes a loophole in v0.2.0 SKILL.md and replaces the example with a stronger demonstration.

Changed

  • SKILL.md — "faithful to the original's structure and intent" → "faithful to the original's argument and intent." The previous phrasing let the skill protect platform-native cadences (staccato fragments, two-sentence quips, three-bullet scaffolds) under the banner of "structure." The skill now explicitly restructures sentences and dissolves formulaic patterns.
  • examples/before.md + after.md — replaced the synthesized marketing-paragraph example with a real LinkedIn post strip. Demonstrates pattern dissolution on competently written but heavily patterned prose, where the issue is cadence rather than empty filler. ~40% word reduction without losing the argument.

See CHANGELOG.md for details.

v0.2.0 — dissolve-ai Strip mode

Choose a tag to compare

@word-ful word-ful released this 17 May 10:56

Default behavior is now Strip mode: paste a draft, get the cleaned draft back. No commentary, no findings, no grade.

Changed

  • Strip mode is the default output.
  • The previous Review-mode findings report is now opt-in via --report.
  • Three action types per flagged passage: Delete, Collapse, Placeholder ([NEEDS SPECIFIC: ...] for passages the skill cannot replace without inventing facts).

Why

A report is a lecture the user did not ask for. The stripped draft is the deliverable. Power users who want the diagnostic can still get it with --report.

See dissolve-ai/README.md for usage.

v0.1.0 — dissolve-ai (review mode)

Choose a tag to compare

@word-ful word-ful released this 11 May 15:13

First public release. Adds the dissolve-ai skill — Review mode for catalogued AI writing patterns. See dissolve-ai/README.md for usage.