ContextQMD
Back to Chromium

Discovery & Batch Selection

agents/projects/code-health/hub/references/discovery_and_batch_selection.md

153.0.7992.31.2 KB
Original Source

Discovery & Batch Selection

Follow this workflow to identify candidates to process.

Steps

  1. Discovery: Run the unified discovery script candidate_finder.py and pass the path to the skill's plugin script:

    bash
    python3 agents/projects/code-health/hub/scripts/candidate_finder.py find --plugin agents/projects/code-health/<skill_name>/scripts/find_candidates.py

  2. Present Candidates/Batch: Output the candidate details or candidate batch to the user. Format the output to list:

    • Batch Directory / File Path: [Directory or File]
    • File / Candidate Count: [Count]
    • Files / Candidate Details: [List of files or details such as Name, Owners, Expiry, Source/XML Paths, Issue Lines, Reason, etc.]
    • [Optional Skill-Specific Metadata]: (Include any additional relevant details, e.g., imports found, warnings, banned symbols, etc.)

  3. Proceed Confirmation: Prompt the user using ask_question. The choices must include the package name for the "Yes" action and "No, discard this batch". If the user rejects the batch, repeat this workflow to find the next candidate batch while excluding the rejected paths.