Skip to content

Two sessions wrote the same tool twice in one day; the answer was an open PR the whole time #3065

Description

@gHashTag

Measured

Twice in one session I started work another session had shipped or had open.

mine theirs when it was visible
a read-only fan-out over workflows, hooks, python gates and Rust --gate commands #3049 tri gates quiet + tri gates empty — two of my four lenses an open PR the whole time I was scanning
a paragraph about a dangling §N cross-reference, and a section confessing one #3056 tri skill refsevery cross-reference, and whether it resolves merged to master while my branch was open on the subject

Neither was hard to find afterwards. Both were free to find before, and the reason I did not is that it takes four commands — gh pr list, gh issue list, git log origin/master, and a grep over the skill files — none of which is habit.

The cost was not theoretical: roughly half of a 29-agent fan-out duplicated ground gates quiet already covers.

Proposed: tri topic <keywords…>

One command over four sources — open PRs, open issues, the last N commits on the base branch, and every SKILL.md section title. A row matches on any keyword; rows are ordered by how many distinct keywords they carry, because the row carrying two of your three words is the one to open.

Verified retroactively on the two collisions above:

$ tri topic gate absent quiet
  rows searched   694
  rows matching   468
  by keyword count  4 with 2, 464 with 1

    2  recent commit  4e27cccac feat(tri): gates quiet -- the steps whose pass survives ... (#3049)
    2  skill section  ci-gates §472. "Cannot check" is not "absent"
    ...
$ tri topic cross-reference renumber
  rows matching   2
    1  recent commit  4ee5b5f54 feat(tri): skill refs -- every cross-reference, and whether it resolves (#3056)
    1  recent commit  519d1891a tri skill renumber: move appended sections ... (#3052)

Both collisions come out at the top of their query.

What it will not do, stated up front

It reads titles and headlines, not bodies or diffs: a PR whose title does not name its subject is invisible here, and the ordering is a word count, not an understanding. 468 of 694 on a three-word query is why the distribution is printed rather than a threshold applied — a cutoff would need a number nobody has measured.

And it refuses when gh cannot answer. "Nobody else is working on this" and "I could not ask" are the same empty list, which is the defect this repository has spent the week naming (#3025, #3045, #3059, #3061, #3063).

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions