Repository navigation
feat: current-generation model guides, freshness checks, and repositioning - #9
Merged
Merged
Conversation
Add dedicated migration guides for Claude Fable 5.1, Opus 5.5, Sonnet 5.5, Haiku 5.5, the GPT-6 family, and Gemini 3.8 Flash. Each guide carries a Last verified date and a Source line. Add references/models/INDEX.md as the registry: model, guide, status, upgrade path, and freshness rules. Mark the Opus 4.8, Opus 4.6, Sonnet 4.5, and GPT-5.5 guides as legacy layers.
Map each error or symptom after a model swap to its cause and fix, with the exact provider error text where documented. Extend the Anthropic, OpenAI, and Gemini provider checklists with the current-generation changes.
Fail validation when a guide lacks a Model ID or Last verified line, when a current guide lacks a Source line, when a legacy guide lacks a Status line, or when a guide is missing from INDEX.md. Require references/breaking-changes.md.
Add scripts/check_freshness.py, which lists dated guides older than 30 days, and a weekly workflow that opens or updates an issue with the report.
max_tokens was 512, which truncates replies on models where thinking is always on and counts toward the cap. Default to 4096 and accept an optional per-scenario request block for effort and max_tokens. Treat a refusal stop reason as a failed case and warn on max_tokens stops. Point the bundled scenarios at claude-opus-5-5 with its list price, and cover the request shape with fake-client tests.
Start model lookup at references/models/INDEX.md, re-fetch provider docs when a guide is stale or the target is not indexed, and point to the breaking-changes lookup. Name the current models and the common 400 errors in the skill description so agents select it for those cases.
Replace the Opus 4.8 launch framing with an evergreen pitch, the exact errors seen after a model swap, a supported-models table, and a sample leaderboard taken from the simulator. Redraw the social preview.
Add .claude-plugin/plugin.json and a single-plugin marketplace so Claude Code users can run: /plugin install modelport --marketplace forkadarshp/MPort The repo root stays the skill, so npx skills add still finds one skill.
Title each current guide "Migrating to <model>: breaking changes and fixes" and add a one-line summary, so the page title and excerpt match what people search for.
Add the plugin install line, per-model prompts, a comparison with the provider guide and an unassisted agent, and FAQ entries phrased as the errors and questions people search for. Update example prompts to current model IDs and drop a roadmap item that already shipped.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Thirteen atomic commits, meant to be rebase-merged.
Newer-model support
references/models/INDEX.md(registry, upgrade paths, freshness rules). Older guides are marked legacy.references/breaking-changes.md: error or symptom to cause to fix, across providers.max_tokens=512, which truncates on models with always-on thinking. Now 4096 by default, with optionaleffortandmax_tokensper scenario and refusal handling.Reach and wording
/plugin install modelport --marketplace forkadarshp/MPort._config.ymlso GitHub Pages can serve each guide as its own page. Pages is not enabled by this PR.Verification
python3 scripts/validate_skill.py .passesnpx markdownlint-cli2 "**/*.md"reports 0 issuespython3 -m unittest discover -s harness/tests: 13 tests passclaude plugin validate .passesnpx skills add . --listfinds one skillNotes