agentleFS
DanMcInerneyDanMcInerney/architect-loop/skills/to-issues/SKILL.md

to-issues

Skill626 starsChanged 4 months agoMITView on GitHub

Use when decomposing a hardened spec into tracker issues: cutting tracer-bullet vertical slices through every layer, ordering structural work first with blocking edges, maximizing the disjoint parallel frontier, publishing producer/consumer interface contracts, and attaching a compact change-skeleton to each issue. Invoked only from an explicit architect dispatch — by the harden strategist subagent in `/architect`, or directly by the orchestrator in `/architect-fast` — after the spec is written; never left to description-trigger discovery.

130 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use decision-mapping, modern-go and 10 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.
  • FabienDehopreSkill

    decision-mapping

    Turn a loose idea into a sequenced map of investigation tickets, then drive them to resolution one at a time.

    Also in 7 projects that use this

  • jidechaoSkill

    modern-go

    Modernize Go code by applying version-appropriate idioms and APIs (gofix-style transformations). Scans go.mod for the Go version, then transforms Go source files to use modern patterns—from Go 1.0 through 1.26+. Use when the user says "现代化","现代Go语言", "地道的", "idiomatic", "modernize", "modern-go", "update Go code", "gofix", or wants to upgrade Go idioms.

    Also in 3 projects that use this

  • victorgarciaesgiSkill

    zoom-out

    Tell the agent to zoom out and give broader context or a higher-level perspective. Use when you're unfamiliar with a section of code or need to understand how it fits into the bigger picture.

    Also in 49 projects that use this

  • jidechaoSkill

    loop-it

    Automated issue loop with checkpoint/resume: fetch open GitHub issues → dependency-aware topological sort → implement each issue end-to-end → review with /review-it → document with /note-it → ship with /ship-it → repeat. Persists state to .loop-state.json for crash recovery. Triggers on: loop-it, loop issues, auto implement, 批量实现, 循环实现, 实现所有issue, 恢复循环, resume loop.

    Also in 3 projects that use this

  • simo97Skill

    request-refactor-plan

    Use this skill when user wants to create a refactor plan.

    Also in 18 projects that use this

  • simo97Skill

    edit-article

    The user will invoke this skill to help them edit an article.

    Also in 18 projects that use this

  • +2 more from jidechao
  • +2 more from simo97

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 123 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Look for prefactoring while you read: make the change easy, then make the easy change.

    line 33
  2. Distinctive:

    Avoid specific file paths or code snippets in issue prose — they go stale fast.

    line 54
  3. Distinctive:

    Every issue carries: acceptance criteria.

    line 76
  4. Very specific:

    Write one draft file per issue at docs/runs/<run>/issues/<NN>-<slice>.md (a local run artifact), NN numbered in publish order: structural …

    line 87

“Common” is measured: kept by at least 6 in 10 of the 123 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
<!-- Adapted from mattpocock/skills (MIT). -->

To Issues

Turn the hardened spec into dispatch-ready issues under the tracking issue. The spec already settles the whole plan — this stage does not quiz the human; a hard stop or an oddity escalation is the only thing that opens a question here.

1. Read before you cut

Load the spec, then hold the codebase-design vocabulary (skills/codebase-design/SKILL.md) exactly: module, interface, implementation, seam, adapter, depth, leverage, locality for design; run, tracking issue, issue, slice, frozen check, check-runner, strategist, builder, orchestrator, factory branch, worktree, job report, verdict, ruling, digest, hard stop for the factory. Never substitute component/service/boundary/API for module/interface, or task/ticket for issue — that drift is a defect, not a style choice.

Look for prefactoring while you read: make the change easy, then make the easy change.

2. Structural before behavioral

Structural (prefactoring) issues land first, with blocking edges into the behavioral issues they unblock. A structural issue's checks prove existing behavior stays green — it changes shape, not outcome.

Oddity rule while scoping: a local wart gets a local patch note, not its own issue. A variation that recurs gets a structural issue blocking the behavioral one. One adapter is a hypothetical seam; two adapters make it real — write the structural issue. Three failed fixes on the same point means stop and escalate the design question instead of cutting a fourth fix.

3. Cut tracer-bullet vertical slices

Each issue is a tracer-bullet vertical slice: a narrow but complete path through every layer the change touches, demoable or verifiable on its own — never a horizontal slice of one layer.

Avoid specific file paths or code snippets in issue prose — they go stale fast. Exception: a prototype snippet that encodes a decision more precisely than prose can (state machine, schema, type shape); trim to the decision-rich parts.

4. Maximize the disjoint parallel frontier

Compute readiness from each issue's file-touch set, not only its blocking edges. Issues dispatched in the same wave share no files, migrations, lockfiles, generated artifacts, config, schemas, or other mutable runtime state. A shared-file collision is a decomposition failure — split or resequence; never hand-merge it later.

5. Interface contracts, not future-code citations

An issue that produces a surface another issue consumes publishes an interface contract block: names, parameters, return types, and behavior. Consumer issues cite that block; they never cite the producer's not-yet-written implementation.

6. Body shape

Every issue carries: acceptance criteria; MAY TOUCH / MUST NOT TOUCH file sets; its check path; its raw job-report path; blocked-by and parent edges; a compact change-skeleton (≤30 lines — files, signatures, data flow, invariants; a contract, not a line mandate). Keep the linked check's RUN dependencies (the test files, validators, and scripts its commands execute) out of MAY TOUCH unless the issue explicitly owns them; when it does, name them in the body as check dependencies. Close every issue body with the run marker comment: <!-- architect-run: <run> -->.

7. Draft in publish order; the orchestrator publishes

Write one draft file per issue at docs/runs/<run>/issues/<NN>-<slice>.md (a local run artifact), NN numbered in publish order: structural issues first, blockers before the issues that cite them. Name blocked-by and parent edges by slice name — real issue numbers do not exist yet; the orchestrator maps slice names to issue numbers as it publishes in NN order, so every blocked-by resolves to a real, already-open issue and never a forward reference. Do not post to the tracker yourself — publishing sub-issues, edges, and claims are orchestrator actions (skills/architect/dispatch.md ## Issue conventions). When the orchestrator runs this skill directly (/architect-fast), it publishes the drafts itself in the same order.

Done when every slice in the plan has a draft file — structural-first, frontier-maxed — whose order publishes with no forward references. End your reply with one line: ISSUE DRAFTS: <count> at docs/runs/<run>/issues/.

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.