agentleFS
cheetah-tradecheetah-trade/claude-code-team-kit/skills/backlog-item/SKILL.md

backlog-item

Skill0 starsChanged 29 days agoMITView on GitHub

Turn a request into a backlog item an agent can finish: title, why, checkable acceptance criteria, out of scope, dependencies; check the backlog for duplicates first and split items that are too big. Use when someone asks for work to be filed, when a conversation ends with work left over, or when an item is too vague to start.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use when a request, a review finding or leftover work has to become a task that a person or a Claude Code session can pick up later and …

    line 8
  2. Distinctive:

    Look for duplicates: Search the backlog (issues, the tracker, the department's backlog file) by keywords and by the screen or file the work …

    line 14
  3. Distinctive:

    Facts yourself, decisions from people: Look up what exists in the code and what was decided before.

    line 18
  4. Distinctive:

    No secrets, customer data or private conversations in an item.

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Backlog item

Use when a request, a review finding or leftover work has to become a task that a person or a Claude Code session can pick up later and finish without asking the author. Pair it with grill-before-build (decisions before the item) and department-brief (which backlog it belongs to).

Before writing

  1. Look for duplicates. Search the backlog (issues, the tracker, the department's backlog file) by keywords and by the screen or file the work touches. If the item exists, add the new fact or criterion to it instead of filing a second one. If it was closed, link it and say why this is different.
  2. One department. An item belongs to one backlog. Work for two departments is two items with a dependency.
  3. Facts yourself, decisions from people. Look up what exists in the code and what was decided before; ask the person who asked only about choices. For anything non-trivial, interview first with grill-before-build.

The parts

  • Title: what changes, in a few words ("Refund button on the order page"). Not "Fix bugs" or "Improve billing".
  • Why: one or two lines on the problem and who has it, with a link to the report, finding or decision.
  • Acceptance criteria: three to seven lines anyone can check without asking the author: an input and the visible result, a command and its output, a screen in a given state. Include the unhappy path and where it must work (phone width, both themes, staging).
  • Out of scope: what not to do here, so the agent does not gold-plate and the reviewer does not ask.
  • Dependencies: what must exist first (another item, a decision, access, a human approval) and what waits on it.

Checkable criteria

Not checkableCheckable
"Refunds work well""A refund over the order total is refused with 'Amount exceeds the order total'; nothing is charged"
"The billing page is fast""The billing page shows the first invoice in under 1 second on the staging data set"
"Handle errors""When the payment provider times out, the order stays 'pending' and the page offers Retry"
"Update the docs""The refunds guide on the docs site covers partial refunds; the link checker passes"

Too big? Split it

Signs: more than seven criteria; criteria for different departments; "and" in the title; a decision nobody has made yet; it cannot be finished and reviewed in one session and one pull request.

Split by a result someone can see, not by layer: each slice ships and is reviewed on its own. Example, "Refunds":

  1. full refund from the order page; 2. partial refund with an amount; 3. refund email to the customer (its wording needs approval); 4. refunds in the monthly export. Group them under one epic and note the order. A "backend" item and a "frontend" item that are useless alone are not a split.

Example item

Title: Partial refund on the order page
Why: support refunds part of an order by hand in the payment provider today; mistakes are not visible in the order.
Acceptance criteria:
- An "Amount" field under Refund accepts 0.01 up to the unrefunded total; anything else shows the reason inline.
- After a refund the order shows "partly refunded" and the amount, and the history row names who refunded it.
- A failed refund leaves the order unchanged and says "Nothing was refunded. Retry".
- Works at phone width in light and dark; `npm test` passes.
Out of scope: refund emails (separate item), refunds in the export.
Depends on: "Full refund on the order page" (done).

When filing

  • Filing is a proposal: the department lead orders and approves the backlog. Do not start an item you filed unless you are asked to.
  • No secrets, customer data or private conversations in an item; summarise and link.
  • Report: the item or its link, duplicates you found, any split with its order, and questions still open.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.