fix-findings

Fix findings on specific files

Targeted clean-up of one or more files using Teamscale's existing analysis.

Steps

1. Reject if no files were given. This skill requires at least one file argument. If the user invoked it without one, ask them which files to clean up and stop.

2. Coverage gate. For each file, fetch coverage:

   ts-agent-helper coverage for-file <file>
   

   If ts-agent-helper exits with a non-zero status, stop the skill and surface its stderr verbatim to the user. Do not guess coverage, fall back to other tools, or retry — the user needs to see the real error (typically a setup, network, or configuration problem).

   If any file's coverage is low (no clear universal threshold — use your judgement; below ~50% line coverage is a strong signal), warn the user:

   Coverage on <file> is low (X%). Cleaning up findings without regression tests is risky — refactors may silently break behaviour. Consider running /teamscale:pr-close-test-gaps first, or adding characterisation tests by hand.

   Ask whether to continue anyway. Stop if the user declines.

3. Fetch findings for each file:

   ts-agent-helper findings list <file>
   

   Same rule as step 2: if ts-agent-helper exits non-zero, stop and surface its stderr verbatim. Do not guess findings.

4. Triage and fix.

To understand finding priorities and fixing strategies, read ../../shared/finding-priorities.md

To get additional information about findings (finding type descriptions) read ../../shared/finding-type-descriptions.md

For flagging findings, read ../../shared/finding-flagging.md

5. Summarise. Print the three buckets (fixed / tolerated / skipped) as described in the priorities snippet.