pr-close-test-gaps

Close test gaps on the current branch

Generate tests for methods introduced or modified on the current branch that are not yet covered.

Steps

1. Fetch test-gap information for the current branch:

   ts-agent-helper test-gaps for-pr
   

   The helper prints a one-paragraph resolution banner on stderr telling you which mode it resolved to:

   • resolved as pull request <id> — an open pull request was found for the current branch; test gaps are scoped to that PR.
   • no pull request for branch '<name>'; comparing against base — no PR matched; the current branch is compared against the repository's default branch (origin/HEAD, falling back to master/main).

   Relay the resolved mode to the user in your first response so they know which scope is being acted on.

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

   On stdout the helper returns CSV data with the following columns:

   • Uniform Path: The path containing the untested method
   • Method Name: The name of the untested method
   • Method Region Lines: The lines for the method as [start - end]
   • Test State: Whether this is an untested addition (new method) or untested change (modified method)

2. Decide which gaps to close. Each row in the CSV is a candidate gap. Pick the ones most worth covering. We give no prior ordering; use your judgement based on the change content (e.g. new public APIs, branchy logic, error paths).

3. Detect the test framework and conventions by reading existing test files in the repo. Do not assume a framework — match what the project already uses (unit vs integration vs other; folder layout; naming). If the repo's CLAUDE.md or test directory makes the convention obvious, follow it.

4. Generate tests for the chosen gaps. Run them after generating to confirm they pass.

5. Summarise. Report which gaps were closed (with test file + test name) and which were skipped (with a one-sentence reason).