structured-gist 0.10.0

Changed

• SKILL.md frontmatter description rewritten (794 characters). It names the summary and report presets and the carve-outs. The triggers "bullet this", "condense this", "give me the gist", "break this down", "make this skimmable" and "tighten this up" are removed because they fired on requests that were not outline requests.
• SKILL.md body compressed from 4,448 to about 2,990 words with no rule change. The Stop-hook activation guidance moved to reference/activation-hooks.md and the per-mode caveman table moved to reference/composition.md.
• SKILL.md ## Pre-emit check now lists linter rules R12, R13, R14 and R15, and the connective-clause test is labelled R12. The rule count reads 16 (R1, R2, R4–R17).

Fixed

• Broken headings, empty code spans and a dangling cross-reference left by an earlier find-and-replace in SKILL.md, reference/composition.md and tests/RESULTS.md. Public wording is now attributed to "a separate public-prose wording skill, if you use one".

structured-gist 0.9.1

Fixed

• structured-gist-claims --help no longer crashes with FileNotFoundError. The command now parses its arguments with argparse: an optional file argument (stdin when omitted), -h/--help and --version. Output and exit codes for file and stdin input are unchanged.

structured-gist 0.9.0

First release on PyPI. It covers the unreleased 0.7.0 and 0.8.0 work and the changes below.

Added

• Python package structured-gist on PyPI. It installs the stdlib-only outline linter as the structured-gist-lint command and the structured_gist import package.
• Linter rule R17, advisory only: structured-gist-lint --advisory flags a ↪ explanation leaf that holds 2 or more commas outside code, quotes and parentheses. Advisories never change the exit code. R17 stays advisory until hedge detection exists (#9).
• SKILL.md pre-emit check: a six-step pass over word caps, enumerator case, delimiter tails, arrow rarity, sibling families and line width, run before an outline is sent (#22).
• Smoke-test ratchet on the benchmark renderings: tests/renderings_lint_baseline.json holds per-rule violation counts, and smoke fails when any rule's count rises (#22).
• Tag-push release workflow: tests, version gate, PyPI publish and GitHub release. Zenodo archives each GitHub release under the concept DOI.

Changed

• Benchmark metric is fact_retention_delta, replacing word_count_reduction_pct (0.7.0). Output length is no longer the headline number (#28).
• The block line budget follows the terminal: width auto resolves to the terminal width minus 8, clamped to 32-120, and 64 when no width is known (0.8.0).
• Citation metadata names Dominik O. Mattioli with ORCID 0000-0002-7327-6337.

v0.6.0

The skill now ships one granularity level, standard. The skim and deep levels are archived for future work.

Why

The controlled benchmark scores the skill at standard only, so skim and deep have no current evidence behind them. Earlier in-repo retention was 0.37 at skim, 0.89 at standard and 0.99 at deep.

Changes

• Usage: the command is now /structured-gist [block [width N|auto]|responsive]. A request for skim or deep renders at standard and says the level is archived.
• Archive: the skim and deep examples, their spec text and a three-level self-explain example are in docs/archive/granularity/. A level comes back only with a controlled benchmark run at that level.
• Word counts:
  • The README word count is restated at standard: 190 to 61 words (-68%, list markers not counted).
  • examples/self-explain.md is remeasured with wc -w: 70 to 28 words (-60%).
• Unchanged: linter rules, render modes, and the experimental KG renderer's three prunes.

Install

/plugin marketplace add domattioli/structured-gist
/plugin install structured-gist

Concept DOI: 10.5281/zenodo.22670068

v0.5.0

First stable release of structured-gist. There is no skill or linter behavior change from v0.5.0b2.

Changes since v0.5.0b2

• Status: beta dropped.
• README:
  • New hero image.
  • Install and Usage come before Motivation.
  • The registrar example is regenerated from the v0.5.0b1 rendering and collapsed.
  • Both outline blocks pass the linter.
  • The linter rule count is corrected to 15 everywhere.
• Badges:
  • Release, License, Tests (live), DOI, Awesome Claude Skills.
  • The static Version, Rules, Tests, Plugin and Dependencies badges are removed.
• CI: a new tests.yml workflow runs pytest and smoke.sh on every push and pull request.
• Citation: CITATION.cff now carries the version and release date.

Install

/plugin marketplace add domattioli/structured-gist
/plugin install structured-gist

Concept DOI: 10.5281/zenodo.22670068