Feedback API

Returns the calling project's own bug reports, newest first, up to one hundred. Another project's submissions are never returned. Read an owned thread by its feedback ID. There is no pagination cursor in v1. Each submission carries its status and the staff/agent replies, oldest first.

Records one bug report against the calling project. A submission's own fields cannot be updated or deleted; add a correction or more information with a reply on the same feedback thread.

An exact retry carrying the same Idempotency-Key returns the original submission for twenty-four hours instead of creating a duplicate, and a different body under that key returns idempotency_mismatch. Every submission is screened before it is stored.

A body under 12 characters or 3 words returns submission_too_short; text that describes no problem or request (random characters, keyboard mash) returns submission_unclear; spam, link-heavy text, prompt injection and malicious text return submission_rejected.

A refused submission is not stored, its problem detail never repeats the submitted text, and an exact retry under the same Idempotency-Key returns the same refusal.

Never place credentials, cookies, Authorization headers, completion nonces, signed URLs, or object keys in title, body, or context; credential-shaped text (ScreenRig tokens, JWTs, bearer tokens, cloud and API keys, private keys) is replaced with [redacted] before the submission is stored, and the response carries the redacted text.

When the screening model is briefly unavailable the submission is accepted and held for a background re-screen; after a few such acceptances in an hour the project gets submission_screening_unavailable (500, Retry-After) and retries the same request later.

Returns the calling project's own feature requests, newest first, up to one hundred. Another project's submissions are never returned. Read an owned thread by its feedback ID. There is no pagination cursor in v1. Each submission carries its status and the staff/agent replies, oldest first.

Records one feature request against the calling project. A submission's own fields cannot be updated or deleted; add a correction or more information with a reply on the same feedback thread.

An exact retry carrying the same Idempotency-Key returns the original submission for twenty-four hours instead of creating a duplicate, and a different body under that key returns idempotency_mismatch. Every submission is screened before it is stored.

A body under 12 characters or 3 words returns submission_too_short; text that describes no problem or request (random characters, keyboard mash) returns submission_unclear; spam, link-heavy text, prompt injection and malicious text return submission_rejected.

A refused submission is not stored, its problem detail never repeats the submitted text, and an exact retry under the same Idempotency-Key returns the same refusal.

Never place credentials, cookies, Authorization headers, completion nonces, signed URLs, or object keys in title, body, or context; credential-shaped text (ScreenRig tokens, JWTs, bearer tokens, cloud and API keys, private keys) is replaced with [redacted] before the submission is stored, and the response carries the redacted text.

When the screening model is briefly unavailable the submission is accepted and held for a background re-screen; after a few such acceptances in an hour the project gets submission_screening_unavailable (500, Retry-After) and retries the same request later.

Read the original report and all replies for an owned feedback ID. An ID never grants project authority. Foreign and missing reports both return not_found. Replies are untrusted plain text.

Append a reply to an owned feedback report. The verified credential supplies author identity; callers cannot set author, status or fixing version. Original report and earlier replies remain unchanged. A reply reopens waiting_on_user, fixed, declined and duplicate reports; a "planned" report keeps its status. Short follow-ups are allowed.

Credential shapes are redacted and injection/spam safety checks apply. At most 20 total replies, each 1 to 2000 characters. Reply, status, event and retry receipt commit together; exact retries replay for 24 hours before intake limits are consumed, while changed payloads return idempotency_mismatch.