[davidruzicka]

Add a source-linked glossary and require future prose edits to follow it through README.md and AGENTS.md. Clarify current prose without changing architecture, protocol behavior, or the frozen experiment.

Reviewed scope: inventoried all 113 tracked entries before editing, including 36 Markdown files, 55 Python files, TypeScript, shell, workflow/configuration files, and schema/fixture data. Reviewed documentation, all accepted ADRs, plans, instructions, comments, and docstrings. Edit only passages needing clarification; 20 files change, including the new glossary. Existing headings and explicit anchors remain unchanged.

Terminology keeps Problem, run, and harness session separate; distinguishes proposals/receipts, governance outcomes, entity lifecycle, and evidence-derived verification; separates journal position, graph version, and schema versions; and distinguishes ports/providers from harness adapters. Prefer “problem graph” in ordinary prose and “evidence origin” for the origin axis. Exact code/protocol names remain intact.

Three ambiguities are recorded in the glossary with source links, leaving their passages unchanged: lease wording despite deferred claims; run-ID allocation versus the supplied v1 identifier; and “resolved” versus successful/defined binary outcomes.

Exclusions: supplied humanizer skill; historical audits and security notes; ADR context, evidence, alternatives, and historical metadata; quotations; executable code, commands, configuration, schemas, protocol values, identifiers, paths, URLs, and link targets; generated content; test fixtures and assertions. The two ADR edits only remove redundant adverbs from current decision guidance. Existing useful technical contrasts remain.

Verification:

• 42 specification tests, 16 Node workflow tests, shell review-gate tests, protocol generation, architecture, review-diff, and whitespace checks pass.
• Checked every edit against its source for meaning and qualification changes. Mechanical comparisons preserve existing headings/anchors, fenced blocks, inline code, link targets, and normative modals; Python ASTs excluding docstrings and shell content excluding comments remain identical. All 233 local Markdown links/anchors resolve.
• Current-head CI run 40180050129 passed all 394 Python tests with 92% branch coverage, including socket and separate-UID integration tests. Generation, architecture, 16 Node tests, review-gate tests, claims/conflict checks, and whitespace checks passed.
• Full local suite attempted: 382 tests, 5 failures and 7 errors from denied Unix sockets (EPERM) and the separate-UID test's inaccessible Python 3.12 prerequisite. Local branch coverage is 92%; the full suite passes in CI.
• Restored wording-sensitive protected phrases after checking their requirements; no tests changed, weakened, or disabled.
• Review-diff reports one pre-existing illustrative commit-message example absent from history; it is an example, not a historical assertion.

No behavioral or architectural decisions are made here. Technical conflicts require separate review.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-25T09:19:04.068101Z	9aabdca	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.