.agents/skills/issue-triage/SKILL.md
Triage GitHub issues for the easyharness repository using the local backlog policy, concrete version milestones, and required rationale comments. Use when reviewing a new issue, backfilling labels on existing issues, revisiting deferred or needs-info backlog items, or deciding whether an issue should stay open, move into a concrete version milestone, or close as not planned.
npx skillsauth add catu-ai/easyharness issue-triageInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Use this skill only for easyharness repository backlog work. This skill
package is the self-contained operational contract for how this repository
triages GitHub issues. Do not rely on external docs to recover the rules.
Keep the backlog legible without adding a heavy project board or a large label taxonomy.
bug, enhancement, documentation, and question when it fits.state/* label on an open issue.v0.x.y only when the issue is
truly accepted into that release scope.state/* label or concrete milestone as "triaged".Use these labels for reviewed issues that are still open but not yet tied to a concrete version milestone:
state/accepted
state/needs-info
state/deferred
If an issue moves into a concrete milestone, remove the state/* label rather
than keeping both.
Use milestones for real version intent, not vague release buckets.
v0.x.y, v0.y.z, v1.x.yA milestone means "this issue belongs to the intended scope of that version"
and should be more specific than state/accepted. It does not by itself mean
the release is ready to cut, nor does it define the release cadence or quality
bar. Those release-policy questions remain separate work, currently tracked by
issue #87.
bug, enhancement, documentation, or question.state/* label, a concrete version
milestone such as v0.x.y, or a close-as-not-planned outcome.state/deferred or state/needs-info, read the earlier
rationale comment first and update the state only when the earlier reason no
longer holds.Whenever an issue is first triaged or later changes triage state, leave a short comment that records:
This comment is required when:
state/* label for the first timestate/* label to anotherstate/* label into a concrete milestoneKeep the comment short and specific. Future backlog sweeps should be able to answer two questions from it:
Example shapes:
Triaged as state/deferred. The idea still looks worthwhile, but the current
repository surface is still moving and I do not want to lock in the wrong
shape yet. Revisit after more dogfooding or when adjacent workflow surfaces
settle.
Triaged as state/needs-info. I do not yet have enough evidence about the user
impact and the preferred UX shape. Revisit once there is a concrete workflow
example or a sharper acceptance target.
Triaged into milestone v0.x.y. This belongs in the next concrete release scope
because it improves an already-shipped maintainer workflow without widening
scope into broader release-policy work.
When revisiting backlog issues:
state/deferred or state/needs-info to state/accepted or a
milestone only when the earlier blocking reason no longer appliesIf the earlier rationale still stands, leave the issue in place and update the comment only when a new fact materially changes the story.
easyharness-managed metadata or move
it into assets/bootstrap/ unless the repository explicitly decides to ship
it later.documentation
Use when acting as a dedicated reviewer subagent for one assigned harness review slot in an existing review round and you need to inspect the change, write structured findings, and submit them through `harness review submit`. This skill is only for reviewer subagents, not for the controller agent.
data-ai
Create or update a tracked harness plan once the direction is clear enough to execute. Use this when work needs a self-contained plan that a future agent can complete from the repository alone, without relying on discovery chat or hidden session memory.
testing
Use when a tracked harness plan has been approved and the controller agent should drive implementation, review, archive closeout, publish/CI/sync evidence work, and merge-readiness follow-up until the archived candidate is genuinely ready to wait for merge approval. This is the main controller skill for day-to-day execution after approval.
testing
Run interactive, Socratic discovery before planning or execution when the objective, boundaries, tradeoffs, success criteria, size, or workflow direction are unclear, or when archived work may need to reopen. Do not use for casual Q&A, simple repo orientation, status checks, or already-approved execution.