firebase-tools/lib/mcp/prompts/crashlytics/connect.js

Command-Line Interface for Firebase

"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
exports.connect = void 0;
const prompt_1 = require("../../prompt");
exports.connect = (0, prompt_1.prompt)("crashlytics", {
    name: "connect",
    omitPrefix: false,
    description: "Access a Firebase application's Crashlytics data.",
    annotations: {
        title: "Access Crashlytics data",
    },
}, async (unused, { accountEmail, firebaseCliCommand }) => {
    return [
        {
            role: "user",
            content: {
                type: "text",
                text: `
You are going to help a developer prioritize and fix issues in their
mobile application by accessing their Firebase Crashlytics data.

Active user: ${accountEmail || "<NONE>"}

General rules:
**ASK THE USER WHAT THEY WOULD LIKE TO DO BEFORE TAKING ACTION**
**ASK ONLY ONE QUESTION OF THE USER AT A TIME**
**MAKE SURE TO FOLLOW THE INSTRUCTIONS, ESPECIALLY WHERE THEY ASK YOU TO CHECK IN WITH THE USER**
**ADHERE TO SUGGESTED FORMATTING**

## Required first steps! Absolutely required! Incredibly important!

  1. **Make sure the user is logged in. No Crashlytics tools will work if the user is not logged in.**
    a. Use the \`firebase_get_environment\` tool to verify that the user is logged in.
    b. If the Firebase 'Active user' is set to <NONE>, instruct the user to run \`${firebaseCliCommand} login\` 
       before continuing. Ignore other fields that are set to <NONE>. We are just making sure the
       user is logged in. 

  2. **Get the app ID for the Firebase application.** 
    a. **PRIORITIZE REMEMBERED APP ID ENTRIES** If an entry for this directory exists in the remembered app ids, use the remembered app id 
       for this directory without presenting any additional options.
       i. If there are multiple remembered app ids for this directory, ask the user to choose one by providing 
          a numbered list of all the package names. Tell them that these values came from memories and how they can modify those values.
    b. **IF THERE IS NO REMEMBERED ENTRY FOR THIS DIRECTORY** Use the app IDs from the \`firebase_get_environment\` tool. 
       i. If you've already called this tool, use the previous response from context.
       ii. If the 'Detected App IDs' is set to <NONE>, ask the user for the value they want to use.
       iii. If there are multiple 'Detected App IDs', ask the user to choose one by providing 
            a numbered list of all the package names and app ids.
    c. **IF THERE IS A REMEMBERED VALUE BUT IT DOES NOT MATCH ANY DETECTED APP IDS** Ask if the user would like to replace the value with one of
       the detected values.
       i. **Description:** A valid app ID to remember contains four colon (":") delimited parts: a version 
          number (typically "1"), a project number, a platform type ("android", "ios", or "web"), 
          and a sequence of hexadecimal characters. 
       ii. Replace the value for this directory with this valid app id, the android package name or ios bundle identifier, and the project directory.
    c. **IF THERE IS NO REMEMBERED ENTRY FOR THIS DIRECTORY** Ask if the user would like to remember the app id selection
       i. **Description:** A valid app ID to remember contains four colon (":") delimited parts: a version 
          number (typically "1"), a project number, a platform type ("android", "ios", or "web"), 
          and a sequence of hexadecimal characters. 
       ii. Store the valid app id value, the android package name or ios bundle identifier, and the project directory.

## Next steps

Once you have confirmed that the user is logged in to Firebase, confirmed the
id for the application that they want to access, and asked if they want to remember the app id for this directory, 
ask the user what actions they would like to perform. 

Use the following format to ask the user what actions they would like to perform:

  1. Prioritize the most impactful stability issues
  2. Diagnose and propose a fix for a crash

Wait for their response before taking action.

## Instructions for Using Crashlytics Data

### How to prioritize issues

Follow these steps to fetch issues and prioritize them.

  1. Use the 'crashlytics_get_top_issues' tool to fetch up to 20 issues.
    1a. Analyze the user's query and apply the appropriate filters.
    1b. If the user asks for crashes, then set the issueErrorType filter to *FATAL*.
    1c. If the user asks about a particular time range, then set both the intervalStartTime and intervalEndTime.
  2. Use the 'crashlytics_get_top_versions' tool to fetch the top versions for this app.
  3. If the user instructions include statements about prioritization, use those instructions.
  4. If the user instructions do not include statements about prioritization, 
  then prioritize the returned issues using the following criteria:
    4a. The app versions for the issue include the most recent version of the app.
    4b. The number of users experiencing the issue across variants
    4c. The volume of crashes
  5. Return the top 5 issues, with a brief description each in a numerical list with the following format: 
    1. Issue <issue id>
        * <the issue title>
        * <the issue subtitle>
        * **Description:** <a discription of the issue based on information from the tool response>
        * **Rationale:** <the reason this issue was prioritized in the way it was>
  6. Ask the user if they would like to diagnose and fix any of the issues presented

### How to diagnose and fix issues

Follow these steps to diagnose and fix issues.

  1. Make sure you have a good understanding of the code structure and where different functionality exists
  2. Use the 'crashlytics_get_issue' tool to get more context on the issue.
  3. Use the 'crashlytics_batch_get_events' tool to get an example crash for this issue. Use the event names in the sampleEvent fields.
    3a. If you need to read more events, use the 'crashlytics_list_events' tool.
    3b. Apply the same filtering criteria that you used to find the issue, so that you find a appropriate events.
  4. Read the files that exist in the stack trace of the issue to understand the crash deeply.
  5. Determine possible root causes for the crash - no more than 5 potential root causes.
  6. Critique your own determination, analyzing how plausible each scenario is given the crash details.
  7. Choose the most likely root cause given your analysis.
  8. Write out a plan for the most likely root cause using the following criteria:
    8a. Write out a description of the issue and including
        * A brief description of the cause of the issue
        * A determination of your level of confidence in the cause of the issue using your analysis.
        * A determination of which library is at fault, this codebase or a dependent library
        * A determination for how complex the fix will be
    8b. The plan should include relevant files to change
    8c. The plan should include a test plan for how the user might verify the fix
    8d. Use the following format for the plan:

    ## Cause
    <A description of the root cause leading to the issue>
      - **Fault**: <a determination of whether this code base is at fault or a dependent library is at fault>
      - **Complexity**: <one of "simple", "moderately simple", "moderately hard", "hard", "oof, I don't know where to start">
    
    ## Fix
    <A description of the fix for this issue and a break down of the changes.>
      1. <Step 1>
      2. <Step 2>

    ## Test
    <A plan for how to test that the issue has been fixed and protect against regressions>
      1. <Test case 1>
      2. <Test case 2>

    ## Other potential causes
    1. <Another possible root cause>
    2. <Another possible root cause>
      
  9. Present the plan to the user and get approval before making the change.
  10. Only if they approve the plan, create a fix for the issue.
    10a. Be mindful of API contracts and do not add fields to resources without a clear way to populate those fields
    10b. If there is not enough information in the crash report to find a root cause, describe why you cannot fix the issue instead of making a guess.
`.trim(),
            },
        },
    ];
});