accountable-review 1.3.0

This release brings major changes to the content of a Review Map, especially a new section called Context. Its aim is to give the reviewer the bits of information they need to understand the behaviour elsewhere in the codebase that the pull request's changes rely on, such as a sequence of requests, a plugin wrapping a core method, or a domain term a setting introduces.

This release also includes a major upgrade to the diagrams inside a Review Map. Each report now picks the type of diagram best suited to what it has to show: a chain for a cause-and-effect path through the code, a convergence for several code paths that must all keep the same invariant, a lifecycle for the states a record moves through and the actions between them, a structure for an entity and its relationships, and a before-and-after request sequence in Context when the change moves which request does what.

What's Changed

• Cut review-map's self-test from 6.5 minutes to under one by @jmbejar in #79
• review-map: topology figures — converge, lifecycle, structure by @jmbejar in #80
• Expand coding-decision checkpoints from one per page to one per departure by @jmbejar in #82
• Add three themes to the review page, chosen by repository config by @jmbejar in #86
• Add logo assets and update branding in examples and README by @jmbejar in #87
• Context: draw the request sequence before and after by @jmbejar in #84
• Context section: repository concepts, the request sequence, and its update rule by @jmbejar in #83
• Add Rust support to review-map skill by @jmbejar in #88
• Port setup-ci tests from shell to Ruby by @jmbejar in #85
• Progress line for interactive review-map runs by @jmbejar in #91
• Add React and Next.js support to review-map skill by @jmbejar in #92
• Move excerpt citation to summary, add --in-entry for Look at entries by @jmbejar in #94
• Read .accountable-review.yml the same way locally and in CI by @jmbejar in #98
• Add --context flag to control Context section visibility by @jmbejar in #93
• Let repositories override the application-code gate with paths.code and paths.skip by @jmbejar in #96
• Clarify citation bar for coding-decision checkpoints by @jmbejar in #99
• Calibrate five more judges against discourse-43002 by @jmbejar in #100
• Regenerate the published examples on every release by @jmbejar in #101
• Give the examples front door Daylight's look by @jmbejar in #102
• Show what a review-map run spends: tokens and the 5-hour window by @jmbejar in #103

Full Changelog: accountable-review--v1.0.5...accountable-review--v1.3.0

accountable-review 1.0.5

Changes since 1.0.1.

The page

• A Review Map no longer points at its own sections with § — sections are named by their linked title, and page-invariants.rb fails a page that writes a section sign outside code (#76).
• <title> is the first line of the emitted page.

Hosts and install

• Codex: review-map installs as a plugin from this repository's own marketplace — codex plugin marketplace add wyeworks/accountable-review (#74).
• Other agents: install the skills with npx skills add; non-Claude hosts get a three-rung ladder for the falsification pass (#69).
• Skill and agent frontmatter is checked by a strict YAML parser (#66). Pi host support was tried and reverted (#68).

Evals

• The synthetic fixture harness is retired. A new end-to-end harness runs the skill against real merged OSS pull requests, checks the page mechanically, and grades it with calibrated LLM judges — the first, for What changed, calibrated on discourse/discourse#43002 (#73).
• setup-ci's slow mutation self-test is dropped (#73).

Docs and repository

• Two example Review Maps published to GitHub Pages (#59, #60); README restructured with a contents list (#57, #58).
• Security policy, code of conduct, issue and PR templates (#62–#65).
• CLAUDE.md trimmed under the load limit (#77).

Full changelog: 1.0.1...accountable-review--v1.0.5