[byte271]

This is FuckMark v0.3.0. The public CLI stays identity: empty release_transform_registry(), process_text("I do not agree.") unchanged, RELEASE_CLI_ALGORITHM_VERSION remains release-cli-v4. No carrier is product-authorized.

The historical v0.2.0 GitHub tag still applies contractions. Do not treat that tag as this product contract.

After merge: do not tag a pull-request SHA if main is a squash merge. Tag the resulting main commit v0.3.0, then workflow_dispatch with publish_github_release=true. This PR itself never creates tags.

Install (F-01, F-02, F-13, F-14)

• Docs and README install from the tagged GitHub Release wheel and SHA256SUMS.txt (v0.3.0 after that release is published).
• New in-repo installers tools/install/unix.sh and tools/install/windows.ps1 verify that checksum, do not use sudo/winget, do not pull main.zip, do not --force-reinstall, do not create 255 case-symlinks, and do not start the CLI.
• d.q1z.org/mark is documented as not the source of truth. That host is not in this repository (F-06), so the live one-click endpoint stays dangerous until it is redeployed from these scripts.

Release (F-03, F-22)

• Package identity is 0.3.0 / v0.3.0.
• Release Engineering no longer auto-tags, auto-publishes, or deletes merged PR branches on main push.
• Publishing is workflow_dispatch with publish_github_release=true only, against an existing v<pyproject version> tag that already points at the dispatch SHA.
• Checkout uses persist-credentials: false. The job never runs git tag.

Workflows (F-05)

• workflow_dispatch values go through env: and "$INPUT_*" instead of ${{ inputs.* }} inside run: scripts.

CLI / contract / research APIs

• Interactive clipboard copy requires --copy (F-17). Without it, interactive mode prints the result.
• load_product_contract() requires frozen hash 5afd79586f82e31d0d673acbebebf0ac00804cff74b9f644f000bddfd3dc07d1 even when specs/ is absent from a wheel (F-08).
• source_verified_release_transform_registry keeps verified visible-preserving rules instead of dropping them (F-09). Visible lexical promotions still fail closed.
• BayesianDetectorEvidence.confirmatory_score() refuses UNVERIFIED status (F-07). Frozen v1 evidence still stores raw_score with UNVERIFIED compatibility.
• load_bayesian_checkpoint rejects JSON larger than 64 MiB (F-18).
• Tracked spec.md is no longer listed in .gitignore (F-20). Root SHA256SUMS.txt remains spec-only.

Intentionally not in this PR

• F-04 / F-12 product decisions (empty CLI, ASCII domain fail-closed identity).
• F-06 website rebuild (mark.q1z.org is a different repo).
• F-10 transformers pin, F-11 YAML confirmatory loader, F-15/F-16/F-19/F-21/F-23–F-26.

  

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: efb845eb-908a-4a34-bcb9-210b6b0e0b59

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d8910e6c41

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Add the configured binary directory to PATH

When FUCKMARK_BIN overrides the default, the launcher is written to that custom directory, but ensure_path still adds only $HOME/.local/bin (and the fish branch has the same hard-coded path). On a machine where the custom directory is not already in PATH, the installer reports success yet fuckmark --help remains unavailable in a new shell; generate the shell entry from BIN or avoid claiming that the override is supported.

Useful? React with 👍 / 👎.