chore(skills): add /add-model and /validate-model commands

[waleedlatif1]

Summary

• add /add-model skill — adds a new LLM model to apps/sim/providers/models.ts with specs verified against the provider's live API docs (no hallucination)
• add /validate-model skill — audits a model entry (or all models in a provider) against live docs and reports drift, dead capability flags, and unverified fields

Type of Change

• Chore

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	May 6, 2026 5:55pm

[cursor]

PR Summary

Low Risk
Low risk: adds new .claude/commands documentation/automation prompts only, with no production code paths or runtime behavior changed.

Overview
Adds two new Claude command definitions, /add-model and /validate-model, to standardize how contributors add or audit entries in apps/sim/providers/models.ts.

The new docs enforce live-docs verification (including two-source pricing checks), outline which capability flags are meaningful via a provider “consumption matrix,” and require structured verification/validation reports plus linting guidance.

^{Reviewed by Cursor Bugbot for commit fc733bd. Configure here.}

[greptile-apps]

Greptile Summary

This PR adds two new Claude slash commands — /add-model and /validate-model — as Markdown skill files under .claude/commands/. They guide an agent through adding or auditing LLM model entries in apps/sim/providers/models.ts using live web fetches, a capability consumption matrix, and mandatory verification reports.

• add-model.md: Defines a five-step workflow (live fetch → consumption matrix → pattern match → lint → verification report) with hard rules against hallucination, explicit lint-failure recovery, and a full anti-patterns list.
• validate-model.md: Mirrors the audit side — reads existing entries, fetches live docs, runs a per-field checklist, and requires human confirmation before any auto-fix is applied; now correctly references add-model.md's canonical provider URL table instead of duplicating it.

Confidence Score: 5/5

Safe to merge — these are instruction-only Markdown skill files with no runtime code paths.

Both files are documentation/prompt files that guide agent behavior; they contain no executable code and require explicit human confirmation before any change to models.ts is applied. The two findings are wording ambiguities in validate-model.md that could confuse an agent at the margin but do not create incorrect data or break any code path.

validate-model.md — the dangling "update the other" sentence and the toolUsageControl checklist item are worth cleaning up before the commands see heavy use.

Important Files Changed

Filename	Overview
.claude/commands/add-model.md	New skill with thorough hard rules, consumption matrix, lint recovery guidance, and mandatory verification report format; no issues found.
.claude/commands/validate-model.md	New validation skill; two minor instruction-quality issues: a dangling "update the other" sentence (the second table no longer exists) and an ambiguous toolUsageControl checklist item that conflicts with the consumption matrix.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A(["/add-model provider model-id"]) --> B[Step 1: Live-fetch provider docs + pricing]
    B --> C[Step 2: Apply Consumption Matrix]
    C --> D[Step 3: Read sibling entries in models.ts]
    D --> E[Step 4: Insert entry + bun run lint]
    E -->|lint fails| F[Fix syntax, re-run lint, note in report]
    F --> E
    E -->|lint passes| G[Step 5: Print Verification Report]
    G -->|any UNVERIFIED rows| H[Ask user to confirm before merging]
    G -->|all verified| I([Done])

    J(["/validate-model provider model-id"]) --> K[Step 1: Read entries from models.ts]
    K --> L[Step 2: Live-fetch provider docs]
    L --> M[Step 3: Re-grep consumption map]
    M --> N[Step 4: Run per-model checklist]
    N --> O[Step 5: Print Validation Report with source URLs]
    O --> P{Offer to fix?}
    P -->|yes| Q[Print diff → get confirmation → edit models.ts → bun run lint → re-check failed rows]
    P -->|no| R([Done])
    Q --> R

Loading

_{Reviews (2): Last reviewed commit: "chore(skills): add /add-model and /valid..." | Re-trigger Greptile}

[waleedlatif1]

Addressed all three Greptile outside-diff findings in fc733bd:

1. Lint-failure recovery (add-model.md Step 4) — added explicit instructions: read the error, fix the syntax/typing issue in the entry just written (do not delete the entry — it's the work product), re-run lint, note the fix in a 'Lint adjustments' line in the verification report. Never report done with lint failing.

2. o3-pro example (validate-model.md line 99) — replaced the misleading parenthetical. The checklist now uses grok-4.3 as the canonical example (xAI docs explicitly state reasoning_effort is not supported) and warns that some always-reasoning models like OpenAI's o-series DO accept the parameter — verify per-model rather than pattern-matching.

3. Duplicated provider URL table (validate-model.md Step 2) — removed the duplicated table; validate-model.md now references the canonical table in add-model.md Step 1 with a note to update both together if either changes.

[waleedlatif1]

@greptile

[waleedlatif1]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit fc733bd. Configure here.}