[schickling-assistant]

Problem

The typed graph omitted requirement-family information for three supported corpus forms: local A01/R01/T01 IDs were emitted as generic id nodes, bold labels split across lines were not emitted as definitions, and legacy labels such as **R01.** retained punctuation in the ID. In addition, template examples such as req: <ID> were emitted as real requirement traces.

These gaps prevented downstream consumers from replacing regex-based Markdown scans without losing definitions or reporting placeholder IDs.

Goal

Make graph nodes and reference resolution recognize the requirement definition forms already present in the corpus, while keeping template placeholders out of req_trace references.

Decisions

• One source-located structured-ID parser now feeds both graph-node construction and reference resolution, so definition grammar cannot drift between the two graph surfaces.
• A trailing legacy period is normalized away only from the ID token; dots inside semantic namespaces remain valid.
• Multiline labels may continue across prose lines, but not across blank lines or fenced-code boundaries.
• Angle brackets are no longer accepted in req: targets. Real IDs and repo-relative path#ID targets remain unchanged.

Verification

Focused graph integration tests:

$ nix develop ../.. -c cargo test --test intent_check graph_json
running 4 tests
....
test result: ok. 4 passed; 0 failed

Full crate suite:

$ nix develop ../.. -c cargo test --locked
10 unit tests passed
15 integration tests passed
0 doc tests failed

The new fixture proves local family kinds, multiline labels, legacy dotted labels, resolved citations to those definitions, and suppression of angle-bracket placeholders.

Complexity

No new module or dependency. The change consolidates two previously separate definition parsers into one shared parser.

Concerns

The graph intentionally treats structured bold ID labels outside requirements.md as definitions too; this is existing graph behavior, now consistently shared with reference resolution.

Friction & bottlenecks

The host Rust toolchain could not link libiconv; running through the repository's Nix development shell provided the declared build environment. No persistent tooling issue identified.

Follow-ups

None.

References

Supports the downstream typed-graph migration in schickling/dotfiles#3596.

Posted on behalf of @schickling

field	value
agent_identity	mbp2025.direct.omp.mjwjfku9
session	unknown
agent_persona	generalist
agent_supervisor	unavailable
agent_tool	OMP
agent_tool_version	18.2.11
agent_runtime	OMP 18.2.11
agent_model	openai-codex/gpt-5.6-sol
worktree	intent/2026-09-27-req-id-definition-graph
tooling_profile	dotfiles@2df08b9-dirty

[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 2026-09-27T10:04:46.606350Z	b3720d6	Draft marked ready

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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b3720d664d

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".