Clarify migration request limits and preset recommendations

skills/migrate-sonar-to-agent-api/references/models-and-presets.md

@@ -15,11 +15,11 @@ Do not invent slugs: there is NO `perplexity/sonar-pro` slug.
 |---|---|---|
 | `sonar` | `preset: "fast"`, or `model: "perplexity/sonar"` + `tools: [{"type": "web_search"}]` | `fast` is the documented starting point. Output is not byte-identical and pricing differs - A/B compare quality and cost before cutover. |
 | `sonar-pro` | Two options - pick per the heuristic below | Option 1, documented starting point: `preset: "fast"`. Option 2, model-slug substitution: `model: "perplexity/sonar"` + web_search. |
-| `sonar-reasoning` | `preset: "low"`, or a current reasoning model + web_search | Retired; do not carry the name forward. Verify the model's supported reasoning efforts and note the substitution in your migration notes. |
+| `sonar-reasoning` | `preset: "low"`, or a current reasoning model + web_search | This skill's recommendation for the retired model, not a mapping specified by the migration guide. Verify the model's supported reasoning efforts and note the substitution in your migration notes. |
 | `sonar-reasoning-pro` | `preset: "low"`, or a current reasoning model + web_search + a generously sized `max_output_tokens` | `low` is the documented starting point; evaluate heavier presets separately. |
 | `sonar-deep-research` | `preset: "high"` | The current migration guide recommends `high` for deep research. |
 
-These are starting points from the [migration guide](https://docs.perplexity.ai/docs/agent-api/migrate-from-sonar/how-to), not guarantees of equivalent behavior or cost.
+The [migration guide](https://docs.perplexity.ai/docs/agent-api/migrate-from-sonar/how-to) supplies the documented preset starting points above; alternative model choices and the `sonar-reasoning` row are this skill's recommendations. None guarantees equivalent behavior or cost.
 
 ### The sonar-pro two-option rule
 

skills/migrate-sonar-to-agent-api/references/request-mapping.md

@@ -128,7 +128,7 @@ Only remove the feature if none of these fit, and call the removal out in the mi
 
 | Limit | Value / behavior |
 |---|---|
-| Document request body | <= 72 MiB including base64, prompts, and other fields; decoded inline documents have a separate 50 MiB combined limit |
+| Native Responses request body | Up to 72 MiB, subject to a lower configured limit, including prompts, base64, and all other fields; decoded inline documents have a separate 50 MiB combined limit |
 | `search_domain_filter` | <= 20 entries, scheme-less, <= 253 chars each |
 | `models` fallback chain | <= 5 |
 | `metadata` | <= 16 keys; key <= 64; value <= 512 |

skills/migrate-sonar-to-agent-api/references/response-and-streaming.md

@@ -113,7 +113,7 @@ Envelope: `{"error": {"message": str, "type": str, "code": int, "param"?: str}}`
 | 400 | `invalid_request`, `invalid_model`, `invalid_parameter`, `invalid_date_format`, `invalid_search_domain_filter`, ... | Unknown field (with `param`), bad model slug, bad filter, missing `max_output_tokens` on `anthropic/*`, reserved function name, expired reconnect cursor. |
 | 401/403 | `forbidden` | Bad or blocked key. |
 | 404 | `not_found` | Unknown response id (also returned for other accounts' ids and `store:false` responses). |
-| 413 | `payload_too_large` | Request body exceeds the applicable limit; document requests allow up to 72 MiB with a separate 50 MiB decoded-inline-document limit. See [request limits](request-mapping.md#limits). |
+| 413 | `payload_too_large` | Native Responses request body exceeds its configured limit (at most 72 MiB). Decoded inline documents have a separate 50 MiB combined limit. See [request limits](request-mapping.md#limits). |
 | 424 | `external_connector_error` | MCP server unreachable. |
 | 429 | `request_rate_limit_exceeded` | Rate limit; retry with backoff; `X-RateLimit-*` headers present. |
 | 5xx | `internal_error`, `bad_gateway`, `upstream_structured_output_failed`, `gateway_timeout` | Retry-worthy. |