Add AI-generated detailed model explanations and metadata for many published models; update skills and audits

[Psuedonerd]

Motivation

• Populate the repository with machine-generated, human-readable detailed explanations for published BNGL models and to supply per-model metadata where missing or incomplete.
• Provide a consistent metadata audit and feature-summary to improve downstream indexing, discovery, and parser-compatibility checks.
• Replace or remove old coder-facing artifacts and update skills used for automated model explanation and metadata auditing to align with the new generation workflow.

Description

• Added AI-generated detailed Markdown explanations for many published models under Published/… (per-model _aigenerated_detailed.md), and corresponding metadata_aigenerated.yaml files for models that lacked curated metadata; several legacy coder explanation files were removed where replaced by the new detailed summaries.
• Added a batch YAML feature audit data/yaml_feature_audit_4.md documenting detected compatibility flags and the bulk changes proposed/inserted into metadata files.
• Updated skill definitions and documentation (skills/explain-model-detailed/SKILL.md and skills/metadata_auditing/SKILL.md) to reflect revised naming, behavior, and the stricter, model-aware explanation/audit rules used to generate the files.
• Preserved existing metadata content where present and inserted or flagged compatibility fields (compartments, functions, anchors, generate-network, etc.) based on active BNGL constructs; generated metadata files are created alongside source YAMLs without overwriting originals.

Testing

• No automated unit or integration tests were modified or executed as part of this change.
• The rollout included content validation steps (presence/consistency checks) performed during generation to ensure each BNGL/YAML pair produced a corresponding Markdown and/or audited YAML file; any ambiguous or parser-sensitive detections were flagged for manual review in the audit report.
• Recommend running the repository's CI schema and simulation-compatibility checks (YAML schema validator, SBML/BNG toolchain) as follow-up to validate generated metadata against project conventions and to exercise simulation actions described in the new files.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

RuleHub/data/yaml_feature_audit_4.md

Line 3 in 0799e33

Audited 30 models and created 30 `metadata_aigenerated.yaml` files. Inserted 239 missing values; 27 items require review.

Complete citations before marking the batch audited

The batch is reported as complete even though 23 of the 30 generated YAML files, including BaruaFceRI2012/metadata_aigenerated.yaml, have no citation mapping, while the other seven retain abbreviated author-year references without the required journal details. This violates the generated-file contract in skills/metadata_auditing/SKILL.md:61-73 and leaves every output without the required complete, verified citation metadata.

---

RuleHub/Published/BaruaBCR2012/metadata_aigenerated.yaml

Line 45 in 0799e33

default_sim_command: ode

Clear defaults when no simulation action exists

For this model, the only active command is generate_network(...); there is no simulate_ode, simulate_ssa, simulate_nf, or simulate call, so default_sim_command: ode falsely advertises a runnable default and is not even one of the command names required by skills/metadata_auditing/SKILL.md:54. The same stale value occurs in Dushek2014 and Kocieniewski2012, while five other new outputs omit this required string-or-null field entirely.

---

RuleHub/Published/Barua2013/metadata_aigenerated.yaml

Line 1 in 0799e33

id: "Barua_2013"

Derive the generated filename from the selected source

When both YAML sources are present, this file copies the higher-priority collaborator source Barua_2013_metadata.yaml but is named as though it came from metadata.yaml. The workflow in skills/metadata_auditing/SKILL.md:19-32 requires Barua_2013_aigenerated.yaml; using the generic name loses provenance and causes a later audit of the generic source to target the same output path. BaruaBCR2012, Blinov2006, Dushek2014, Hat2016, Kocieniewski2012, and Kozer2014 have the same collision.

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".