agentleFS
gonimargonimar/claude-web-studio/skills/backlog/SKILL.md

backlog

Skill4 starsChanged 46 days agoMITView on GitHub

Records an idea, wish or suggestion that nobody has decided to act on as one line in production/backlog.md and stops — captured, not built. Use the instant one surfaces: a stray 'what if…', 'maybe we should…', 'someday…', 'it'd be nice if… (not now)' dropped mid-conversation about something unrelated to the current task; any instruction to remember something for later — 'add to backlog', 'file this', 'note for someday', 'park this', 'log as a future item' — including a gap named without the word backlog ('we have no X', 'there's no Y yet'); a question about the idea list, its oldest or top item, or the periodic review; promoting an idea to /brainstorm, /impact or /feature-spec, parking or closing it. The tell: nothing gets built, decided or estimated this turn, only written down. Not for a request already decided and ready to execute (an assigned story, a requested spec, a ticket closed with passing tests) or a generic question about backlogs as a concept.

36 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use issue, worktree and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • AWSSkill

    issue

    Implement a GitHub issue end-to-end — branch, code, test, docs, and commit.

    Also in 3 projects that use this

  • MicrosoftSkill

    worktree

    Create, reuse or clean up a dedicated git worktree for working on a pull request or a private branch. Use from any PR-scoped workflow that needs an isolated checkout to keep current code and build output untouched.

    Also in 3 projects that use this

  • AWSSkill

    retro

    Conduct a retrospective on a coding session.

    Also in 4 projects that use this

  • BlockSkill

    plan

    Draft a TDD, PRD, or lightweight plan in the conversation without committing it.

    Also in 4 projects that use this

  • Hugging FaceSkill

    implement

    Make tests pass. Invoke after /write-tests produces failing tests.

    Also in 4 projects that use this

  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 35 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    After the "write" answer: touch .claude/.write-consent (rule 7 — the consent-guard hook checks the marker), then write.

    line 18
  2. Very specific:

    Read production/backlog.md and take the next I-NNN.

    line 35
  3. Very specific:

    Write gate and commit gate (docs: backlog I-NNN <title>).

    line 43
  4. Very specific:

    After the hand-off returns, mark the entry → promoted YYYY-MM-DD: F-NNN (or the ADR / impact verdict) and tick it [x], behind the write …

    line 69

“Common” is measured: kept by at least 6 in 10 of the 35 edited versions of this file we could read. “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 — ideas are recorded, not executed

Reply in the project conversation language (CLAUDE.md → Language); code, identifiers, paths and commit messages stay in English.

Template .claude/docs/templates/backlog.md; the roadmap format (.claude/docs/templates/roadmap.md) reserves production/backlog.md for ideas before their spec and production/decisions.md for owner decisions. In the commands below, <hooks> is .claude/hooks/ in copy mode and ${CLAUDE_PLUGIN_ROOT}/hooks/ in plugin mode.

Write gate (every mode below that changes the file): render the new or changed entry in the chat, then "May I write production/backlog.md?" — one AskUserQuestion: write (Recommended) · adjust the wording · not now. After the "write" answer: touch .claude/.write-consent (rule 7 — the consent-guard hook checks the marker), then write. Then one commit gate: docs: backlog <what> staging exactly that file, on the branch git-workflow's documents lane prescribes. Record the gate before asking — <hooks>session-state.sh set Gate "/backlog <mode>: commit?" — and clear it after the answer (<hooks>session-state.sh set Gate "—").

Not initialised — no .claude/docs/ and no production/ → BLOCKED (not initialised — run /init), nothing written. An initialised project without production/backlog.md gets it created from the template at the first write.

Phase 0: What is an idea (rule for every skill and the main conversation)

A musing — "what if we…", "maybe we should…", "it would be nice…", or their equivalents in the conversation language, said while something else is in progress — is an idea, not an instruction. Nobody has decided anything yet, so nothing is implemented, drafted or planned in the same turn (CLAUDE.md principle 2; coordination-rules rule 11). The main conversation names it ("that is an idea, not part of the current story") and offers exactly this skill: /backlog add "<the idea in the user's words>". A proposal with a decision behind it ("add Redis for the session store") is /impact's business, not the backlog's; when unsure, record it here — promoting later costs one command, implementing an undecided idea costs a story.

Phase 1: add "<idea>"

  1. Read production/backlog.md and take the next I-NNN.
  2. Draft one entry: ### I-NNN · <title in ≤ 8 words> 💡, one line recorded YYYY-MM-DD, source: <conversation | user | incident INC-NNN | audit>, and the idea in one or two sentences in the user's words. Add the area (🏷 backend/frontend/game/infra/docs) when it is obvious. No analysis, no options, no estimate: that is what promotion is for.
  3. A proposal passed to add (a concrete change with a decision behind it) is recorded as asked — the backlog never refuses — with one line saying that a concrete change is /impact's business; the next step then offers /backlog promote I-NNN → /impact.
  4. Write gate and commit gate (docs: backlog I-NNN <title>).
  5. Stop here. The recorded idea never becomes a task in this turn. If a story or skill was in progress, the hand-off returns to it ("back to /dev-story S-NNN (Recommended)").

Phase 2: review (weekly, or when /help shows Backlog: N ideas, oldest N days)

  1. Render a table of open ideas: ID · title · recorded · age · area · marker (🅿 parked). Flag ideas older than 60 days without a decision.
  2. Ideas that left sideways. An idea can leave the backlog without promote — the user ran /impact or /feature-spec on it directly — and the entry then stays open for weeks while the work is already done. An open entry whose text matches a spec, ADR or impact verdict written later is that case: name it and offer to close it with the link.
  3. One AskUserQuestion per review, never per idea: promote one (name it) (Recommended when something is older than 30 days) · park/close some · nothing now.
  4. Apply the answer through the matching mode below. With a write answer, also set last-review and Reviewed: to today (the template header; /help reads last-review), behind the write gate.

Ideas are never silently deleted: close keeps the line with [x] and the reason.

Phase 3: promote I-NNN · park I-NNN · close I-NNN <reason>

promote I-NNN — the idea goes through the pipeline it skipped.

  1. Decide the route from the entry, not from its wording:
    • vague, no clear user or scope → /brainstorm "<title>";
    • a concrete change to code, architecture, security or deployment → /impact "<idea>" (its verifiers decide);
    • a user-visible feature the product spec allows → /feature-spec "<title>", it receives F-NNN;
    • a technical choice → /architecture-decision "<title>".
  2. Hand off with one AskUserQuestion: the route chosen in step 1 (Recommended) · the other plausible routes · not now.
  3. After the hand-off returns, mark the entry → promoted YYYY-MM-DD: [F-NNN](../docs/specs/features/…) (or the ADR / impact verdict) and tick it [x], behind the write gate ("May I update production/backlog.md?") and the commit gate. The promotion counts only once that line exists.

park I-NNN adds 🅿 and a date (and the reason, when given). close I-NNN <reason> ticks it [x] with the reason. Both go through the write gate and the commit gate.

Nothing here ever creates a story or writes code.

Verdict: COMPLETE (I-NNN recorded) | COMPLETE (review: N open, M promoted) | COMPLETE (I-NNN promoted → …) | BLOCKED (not initialised — run /init). Next step — one AskUserQuestion: return to the interrupted work (Recommended when something was in progress) · /backlog promote I-NNN (Recommended when a proposal was just recorded and nothing was in progress) · /backlog review · stop here.

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.