AGENT SKILL

triage

Backlog hygiene sweep - stale items, duplicate pairs, missing acceptance criteria, unblocked-but-unassigned work, and milestone burn, delivered as a report first with only user-approved actions applied. Works on GitHub Issues in tickets mode and on PLAN.md checkboxes in document mode. Use when the user asks to triage, groom, or clean up the backlog or the plan's open tasks.

License 许可
MIT
最近更新
2026-09-28
Stars
★ 2
Forks
0

01 前言 Front Matter

name
triage
description
Backlog hygiene sweep - stale items, duplicate pairs, missing acceptance criteria, unblocked-but-unassigned work, and milestone burn, delivered as a report first with only user-approved actions applied. Works on GitHub Issues in tickets mode and on PLAN.md checkboxes in document mode. Use when the user asks to triage, groom, or clean up the backlog or the plan's open tasks.
argument-hint
[notes]

02 资源文件 2

03 SKILL.md 正文

/triage — keep the backlog honest

A backlog rots in specific, findable ways. This skill finds them, reports them with evidence, and applies exactly the actions the user approves — nothing more, and nothing silently.

Adjacent skills: /audit docs (read-only drift report; /triage proposes and applies tracker/plan actions) · /ticket (files one new item; /triage grooms the many existing ones).

<!-- acstack:runtime --> Run before the skill's steps — per invocation, not per session (4.36); failures degrade to markdown:

link="$(readlink "$HOME/.claude/skills/health" 2>/dev/null || true)"   # empty = not symlinked
pack="$(dirname "$(dirname "$link")")"   # NEVER trust this unless $link was non-empty
if [ "${link#/}" != "$link" ] && [ -x "$pack/bin/acstack-config" ] && ! "$pack/bin/acstack-config" runtime | grep -q '=off'; then
  "$pack/bin/acstack-config" || true          # resolved keys, with sources
  "$pack/bin/acstack-update-check" || true    # ≤1 fetch/day; silent ONLY if already checked today
  "$pack/bin/acstack-recall" || true          # LEARNINGS.md + bug-class names, capped 3KB
else
  echo "runtime off — proceeding without recall/update-check"
fi

<!-- /acstack:runtime -->

<!-- acstack:principles -->

Operating principles

  • Be direct. Push back in writing when the plan or the user is wrong. No sycophancy.
  • Never delete a decision. Supersede it: ~~old~~ → **Verdict (YYYY-MM-DD):** new call — reason.
  • Never fix, tune, or delete a test or eval case to raise a score. Log the miss honestly and leave the case unchanged.
  • Name exact things: regex patterns, function signatures, model names, before → after numbers. Never "fixed bugs".
  • Attribution: follow the project's attribution setting (default none) — no AI-tool mentions in generated docs, no attribution trailers in commits or PRs. Commit with explicit -m/-F messages only.
  • Config: read .claude/acstack.md at the project root (fall back to ~/.claude/acstack.md) before acting. ## Settings keys override pack defaults; a ## <skill-name> section overrides both. Unknown keys and sections are ignored.
  • Docs: BRIEF.md (frozen seed) / PLAN.md (living plan) / JOURNAL.md (rolling journal). If the repo uses legacy names (PLANNING_PROMPT.md / PLANNING.md / STATUS.md), use those instead — never create both.
  • Recall: if LEARNINGS.md exists at the project root, read it before starting.
  • Conduct: follow the acstack-conduct block in this repo's AGENTS.md — the word is the mode; the user sets the pace.
  • Hackathon lane: if the project's AGENTS.md carries the acstack:hackathon-lane block, only /do changes the repository during the event. Any other skill that would write a tracked file, commit or push says what it would have done and stops; a change that is not a task goes through the lane's operator route. <!-- /acstack:principles -->

One document set. Resolve exactly ONE BRIEF/PLAN/JOURNAL set and name its path in the report's scope line. If more than one candidate set exists — a monorepo, nested products, an apps/* tree each with its own docs — list the candidates and STOP. Never pick one silently: a confident answer about the wrong product is worse than no answer (conduct rule 8).

Stance: report, then apply

The report opens with a one-line verdict — backlog healthy or <N> findings across <M> categories — before any finding. Then the numbered findings, each with its evidence and its proposed action. Then the user picks; only approved actions are executed, and the report closes with the applied-actions list — each action confirmed by name. Nothing is ever silently deleted or closed: every close carries a written reason. Supersede-don't-delete applies to backlogs exactly as it applies to decisions.

Applied actions edit the plan in place, so they collide with other sessions. Closing or reopening a box changes any count derived from those boxes; two sessions closing different items both hand-write the same new total, and git auto-merges the identical edit with no conflict — measured 2026-09-14, leaving the number one out. So: apply the box changes, and re-derive any stored total with the project's own tool rather than editing the digit. State in the applied-actions list that the post-merge total belongs to whoever integrates, since this branch cannot see what another session closed.

Config: stale-days from a ## triage section in .claude/acstack.md (default 30).

Mode: the sweep itself

The sweeps are mode-specific and mutually exclusive — read exactly one, chosen by the resolved tracking config, and not the other:

  • tracking: tickets → references/tickets-mode.md
  • tracking: document (the default) → references/document-mode.md

Root-cause clustering below runs in BOTH modes, after whichever sweep you read.

Root-cause clustering (both modes)

Every sweep above is a local comparison — this item against that item, this box against its acceptance. Clustering is the global one: twelve items that are all one cause is a finding no pairwise duplicate check can see, because no two of them are duplicates.

Run it last, over the whole backlog at once, after the local sweeps have already removed the noise.

A cluster needs a stated cause, not a shared topic. "These five all mention the parser" is a tag. "These five all fail because the tokenizer drops the final field when the input has no trailing newline" is a cause. The test is one sentence: would fixing the named cause close every item in the cluster? If some would survive it, the cluster is wrong — split it or drop it.

The bar, all three required:

  • Three or more items to propose a PARENT. A parent over two items is more structure than it earns. But a same-cause pair is still named, never dropped. Two items sharing a cause are not a duplicate pair — the duplicate sweep keys on overlapping text and will not see them, so a pair that falls below this bar would otherwise vanish between the two sweeps. Report it under Related pairs: the cause, both items, and the evidence, with no parent proposed. Found by shakedown 7, where two items sharing a UTC-boundary cause were invisible to both passes.
  • A cause written as a sentence, naming the mechanism — not a component, a label, or a theme.
  • Evidence per member: the quoted line, error, or acceptance that ties that specific item to that cause. An item nobody can tie to the cause is not in the cluster.

Propose nothing when the backlog is genuinely independent. This is the half that keeps the pass honest: a backlog of unrelated work is the normal, healthy state, and a clustering step that always finds clusters is astrology. Say plainly that no root-cause groups were found and move on — that is a real result, not a failed sweep. Never widen a cause until items fit it; a cause broad enough to cover everything explains nothing.

What the report proposes, per cluster:

  • A parent — in tickets mode a new issue titled by the CAUSE; in document mode a parent task in PLAN.md with the children as its subtasks. Never a parent that merely renames the group.
  • The children redirected, each with the same standardized line so the tree stays navigable: rolled up into <parent> — same cause: <cause>.
  • What stays independent, listed. An item that resisted clustering is information, not a leftover.

Approval-gated exactly like the rest of this skill: the clusters are proposed with their evidence, and nothing is filed, edited, or closed until the user says so. Getting a cluster wrong reshapes someone's whole backlog, so this is the sweep where a wrong proposal costs the most.

Hard rules

  • Findings without evidence (the quoted text, the failing command, the date arithmetic) don't make the report.
  • The report states what was NOT swept (e.g. acceptance commands too expensive to run) — a sweep that implies totality it didn't do is drift about drift.
  • Applied actions are listed at the end exactly as executed — issue numbers, label names, PLAN.md lines — so the JOURNAL entry can cite them.

04 同仓库更多 Skills

AaravChadha
AaravChadha/acstack

why

Decision archaeology - answers why the code, config, or plan is the way it is by searching the record in a fixed order: BRIEF constraints, dated PLAN decisions, JOURNAL entries, then git history. Stops at the first real answer, cites it with a date, and says no recorded rationale rather than inventing one. Use when someone asks why something is this way, why a choice was made, or where a decision is written down.

★ 2
AaravChadha
AaravChadha/acstack

verify

Audit a completion claim made by someone else - another session, agent, or teammate - against the running system rather than the diff. Re-derives what the claim's acceptance demands, runs it, and returns CONFIRMED, OVERSTATED, FALSE or UNVERIFIABLE with the command and its output. Never fixes, never edits the claim or the acceptance. Use when handed a "done", "all green", "merged" or "it works" you did not produce yourself.

★ 2
AaravChadha
AaravChadha/acstack

ticket

Capture a brain-dump as a well-formed work item - verb-first title, acceptance criteria, file paths when known, an out-of-scope line. Files a GitHub issue in tickets mode or appends a numbered PLAN.md task in document mode; unknowns are marked TBD, never invented. Use when the user brain-dumps an idea, bug, or task to capture for later, or asks to ticket, file, or note something as work.

★ 2
AaravChadha
AaravChadha/acstack

ship

Branch-level release with five gates before the act - clean-state, tests, eval-vs-target, docs drift, and attribution sweep; any failing gate stops the release with its output. Then push, and under push: branch-pr open a report-shaped PR wiring the issue-closing reference in tickets mode or PLAN task IDs in document mode. Use when the user asks to ship, release, cut, or open the PR for a feature or branch.

★ 2
AaravChadha
AaravChadha/acstack

secure

Security review with a confidence gate - a finding exists only with a concrete exploit scenario and a high/medium/low confidence rating; everything else goes to a worth-hardening list, never inflated. Sweeps five surfaces: auth gates, secrets hygiene, injection and unsafe sinks, unsafe deserialization/crypto/transport, and LLM tool-use trust boundaries. Reports only, never fixes. Use when the user asks for a security review or to check vulnerabilities, secrets, or auth.

★ 2
AaravChadha
AaravChadha/acstack

retro

Trend review across many sessions - velocity vs planned dates, eval-score trend toward target, failure-category trends from journal triage and investigations, and a status check on every open PLAN risk. Reads JOURNAL/PLAN/eval history, writes a dated retro entry into JOURNAL.md. Use when the user asks for a retro, a weekly review, or a phase-end wrap-up.

★ 2