Changelog

Skill versions track the bundled references/schema.graphql, downloaded from https://ax0.taddy.org/docs/taddy-podcast-api-schema.graphql at each release. Re-copy the docs mirror and bump the version whenever the API changes.

1.0.3 (2026-11-06)

• README: install is now one prompt that asks the agent to download the skill. The per-agent Claude Code, Codex and Cursor marketplace steps are gone. "Layout" is renamed "Architecture".

1.0.2 (2026-11-06)

• Docs mirror and schema.graphql re-copied from taddy.org.
• Schema: the webhook mutations and the WebhookFilterInput / WebhookFilter fields now have descriptions, covering AllowList and BlockList, AND/OR filter matching, and the fact that updateWebhookFilter replaces the whole filter.
• getPopularContent, getTopChartsByCountry and getTopChartsByGenres now allow page 8 at most (was 20).
• Search: the matchBy options are now MOST_TERMS, ALL_TERMS and EXACT_PHRASE (FREQUENCY is gone). search.md adds a section saying that transcript text is not searched.
• Transcripts: the wording says Paid, or Pro or Business, where it used to say Business. taddyTranscribeStatus explains that a transcript can be generated on demand whatever the status.
• ai-tools-mcp-server.md, the intro page and the "Get Started" banners now point to the skill and the MCP server.

1.0.1 (2026-11-05)

• README: rewrote the install instructions. Notes that Claude Code and Codex plugins can auto-update, adds the Codex /plugin step, and replaces npx skills add with a prompt that asks the agent to download the skill.

1.0.0 (2026-11-02)

• Typed client option: SKILL.md now keeps the plain HTTP client for side features and sets up a typed client (GraphQL Code Generator + graphql-request, Node 22+) when podcast data is the product. New references/graphql-client-libraries.md with per-language options, plus scripts/taddy_client_typed.ts, scripts/codegen.ts, and starter scripts/graphql/ fragments and queries. Codegen introspects https://api.taddy.org at build time so generated types always match production.
• Error handling: errors-and-quota.md and SKILL.md now document the HTTP status behind each Taddy error code, the 4xx-vs-200-with-errors[] distinction, the per-minute rate limit, and the QUERY_TOO_COMPLEX cost model. The JS and Python clients surface the Taddy error code on non-2xx responses.
• Clients: endpoint fixed at https://api.taddy.org (no more TADDY_API_URL); all three clients default to a 30-second timeout and throw TaddyError with code: "TIMEOUT".
• Throttling: new scripts/taddy_throttled.js and scripts/taddy_throttled.py wrap query() with a 250 ms pause after each request, so loops, cron jobs and polling stay under the per-minute rate limit. SKILL.md routes that work to them before the first API_RATE_LIMIT_EXCEEDED; errors-and-quota.md points at them from the 429 row.
• Podcast-only sources: the escape hatch and client-libraries page link the podcast-only schema and docs (taddy-podcast-api-schema.graphql, taddy-podcast-api-docs.md) instead of the all-media files.
• Feedback: new references/skill-feedback.md and scripts/taddy_feedback.js. The agent drafts a six-field report (kind, summary, details, operations; skill name and version filled in), the script validates and scrubs it, and only --send after the user's approval posts it with the account headers. Needs the feedback endpoint on taddy.org (spec in proposals/skill-feedback-endpoint.md); until it ships the script says so and nothing is sent. evals/check-refs.sh now syntax-checks every script and runs the redaction fixture.
• SKILL.md: triage by project state (setup / add a query / debug), query decision table, gotchas, reference routing, escape hatch.
• References: full podcast docs mirror (49 pages) plus schema.graphql, and two curated pages, pagination.md and errors-and-quota.md.
• Scripts: taddy_client.js, taddy_client.py, webhook-express.js.
• Manifests: the Claude Code plugin.json now sets privacyPolicyUrl and supportUrl, and the Codex support link points at https://taddy.org/developers/support instead of the terms of service. Both URLs live at the top level of data.json.
• Plugin manifests for Claude Code (.claude-plugin/), Cursor (.cursor-plugin/), Codex (.agents/plugins/), and the Agent Plugins standard (root plugin.json), all generated from data.json by scripts/sync-manifests.py.