Conversation
Remove the daily submission rule and the exact attempt rate from the feedback tool guidance and README. Each keyless job accepts one submission, retries return the original feedback ID, and attempts are described as rate limited. Model the smoke test's 429 on the API attempt throttle response.
Resolve conflicts with Alexandria session feedback, the compact response format, and the tool description budget: - Register firecrawl_feedback for keyless sessions while Alexandria result pointers keep their authenticated availability check. - Accept Alexandria session feedback and keyless job feedback through the same tool, and keep exposing the Search id to keyless callers. - Move the keyless observation rules to the observations parameter so every tool description fits the client description window, and place keyless feedback copy after the routing copy. - Describe keyless feedback as requested in exchange for free keyless access.
…E copy Replace the remaining optional keyless feedback wording in the local keyless startup message and the Search tool README entry.
Restore the authenticated Search id sentence from main next to the keyless feedback sentence so both contracts stay described.
…rong Replace the in-exchange framing with an optional ask tied to a trigger: if a result is wrong, incomplete, blocked, or an error. This covers the Search, Scrape and Parse descriptions, the feedback tool, the keyless instructions, the startup message and the README. The descriptions say "via" so Search stays inside the 2,048-character description cap.
…ions Tool descriptions are read by every session, so authenticated agents should not carry keyless-only guidance. Keyless agents still get the ask in each keyless result's invitation, the feedback tool's description and the keyless instructions.
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
1 issue found across 6 files
Confidence score: 4/5
- In
src/index.ts, failed keyless jobs omit thejobIdfrom the displayed error, so hosted clients may not show users which job failed. Include the ID in the message text.
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="src/index.ts">
<violation number="1" location="src/index.ts:3194">
P2: A failed keyless job throws a UserError whose message is only the API error text; the jobId is passed only as the structured payload. Hosted clients receive the text block (see the convention noted at the `KEYLESS_SIGNUP_FALLBACK_URL` comment, and `executeExchangeCalls`, which embeds requestId in the message for this reason), so the caller cannot obtain the job reference needed to submit `kind: failure` feedback that the keyless instructions advertise. Include the job reference in the message text.</violation>
</file>
Reply with feedback, questions, or to request a fix.
Fix all with cubic | Re-trigger cubic
| if (json?.metadata?.jobId) { | ||
| throw new UserError( | ||
| json.error || `Firecrawl request failed (HTTP ${response.status})`, | ||
| json |
There was a problem hiding this comment.
P2: A failed keyless job throws a UserError whose message is only the API error text; the jobId is passed only as the structured payload. Hosted clients receive the text block (see the convention noted at the KEYLESS_SIGNUP_FALLBACK_URL comment, and executeExchangeCalls, which embeds requestId in the message for this reason), so the caller cannot obtain the job reference needed to submit kind: failure feedback that the keyless instructions advertise. Include the job reference in the message text.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At src/index.ts, line 3194:
<comment>A failed keyless job throws a UserError whose message is only the API error text; the jobId is passed only as the structured payload. Hosted clients receive the text block (see the convention noted at the `KEYLESS_SIGNUP_FALLBACK_URL` comment, and `executeExchangeCalls`, which embeds requestId in the message for this reason), so the caller cannot obtain the job reference needed to submit `kind: failure` feedback that the keyless instructions advertise. Include the job reference in the message text.</comment>
<file context>
@@ -3194,6 +3191,12 @@ async function keylessPost(
if (hints) payload.agent_hints = hints;
throw new UserError(String(payload.message), payload);
}
+ if (json?.metadata?.jobId) {
+ throw new UserError(
+ json.error || `Firecrawl request failed (HTTP ${response.status})`,
</file context>
| if (json?.metadata?.jobId) { | |
| throw new UserError( | |
| json.error || `Firecrawl request failed (HTTP ${response.status})`, | |
| json | |
| if (json?.metadata?.jobId) { | |
| const message = | |
| json.error || `Firecrawl request failed (HTTP ${response.status})`; | |
| throw new UserError( | |
| `${message} Job reference: ${json.metadata.jobId}`, | |
| json | |
| ); | |
| } |
|
@cubic-dev-ai review this PR in full at the current head. Use the API contract in firecrawl/firecrawl#4616 as the source of truth. Keyless feedback is optional, available for Search, Scrape, and Parse, and requires task, assessment, observations, and Parse docClass. The shared tool schema retains optional evidence fields for authenticated compatibility; keyless sessions validate the required fields before a request. Failed jobs retain the full API envelope in text and structured output, including HTTP 200 responses with success: false. Feedback submissions bypass operation eligibility checks but retain trusted hosted caller identity. Please recheck the complete diff and the prior findings against the current implementation. |
@Max17190 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
No issues found across 5 files
Confidence score: 5/5
- Automated review surfaced no issues in the provided summaries.
- No files require special attention.
Requires human review: Expands keyless MCP exposure by adding firecrawl_feedback to anonymous callers and widening the public tool set and search output contract; the rollout, identity, and abuse tradeoffs need human sign-off.
Re-trigger cubic
This branch has not been deployed
1 participant
Why
Let agents discover and optionally submit evidence for keyless Search, Scrape, and Parse using the API contract in #4616.
Summary
firecrawl_feedbacktool to keyless callers, including after operation quota exhaustion, and forward trusted hosted caller identity.Release after the API is deployed and Docs #1414 is published.
Test Plan
success: falseis an MCP error with its job reference in both text and structured output; its failure feedback and retry succeed.