beads-writer
Use when creating or decomposing Beads issues for a project backlog, including new features, bugs, epics, backlog grooming, or converting vague asks into actionable Beads work in a Beads-backed repository.
- License 许可
- MIT
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- beads-writer
- metadata
- { "owner": "tze", "status": "active", "authors": [ "tze", "OpenAI Codex" ], "last_reviewed": "2026-11-06" }
- description
- Use when creating or decomposing Beads issues for a project backlog, including new features, bugs, epics, backlog grooming, or converting vague asks into actionable Beads work in a Beads-backed repository.
- compatibility
- Requires a Beads-backed repository with the `bd` CLI available. Assumes the agent can inspect the target project workspace before creating issues.
02 资源文件 5
- 参考文档agents/openai.yaml313 B
- 参考文档references/fields-and-examples.md7.1 KB
- 参考文档references/quality-checklist.md2.8 KB
- 参考文档references/reconciliation-beads.md3.4 KB
- 参考文档references/workflow.md5.3 KB
03 SKILL.md 正文
Beads Writer
Create high-quality beads issues through repo-grounded research and deliberate field selection. This skill is a routing layer over the Beads operating model described in ../../../../../README.md: issue writing should reinforce that workflow, not compete with it.
Write a concise current packet grounded in the proportional governing contract from work-allocation so a fresh owner can act without the creation chat. Tracking and delivery granularity differ: related existing beads may share one package while retaining their own acceptance.
Use This Skill When
- A user asks to create one or more beads issues for a feature, bug, task, or backlog item
- A user wants a vague ask decomposed into actionable beads or an epic with child beads
- A discovered issue needs a proper bead instead of a quick title-only stub
- A backlog-grooming pass needs better scope, dependencies, or acceptance criteria
Do Not Use This Skill When
- The work already exists as a well-formed bead and should now be implemented: use
beads-worker - The task is to review a whole repository or generate a roadmap from audit findings: use
th-projects(project-review or project-direction subskills) - The task is exhaustive spec-versus-code reconciliation rather than issue writing: use
th-projects(spec-concretization subskill)
Core Rules
- Ground every bead in the target repo's actual code, naming, issue graph, and constraints before writing it.
- Prefer one bead per testable outcome. Split unrelated acceptance criteria.
- Apply the shared
th-projectswork-allocation contract before creating or decomposing implementation work. - Keep accepted outcomes finite and visibly partial until mandatory acceptance is met.
- Use one cross-child closeout path when needed; an existing outcome checklist or report may suffice. Do not create recursive reconciliation generations or split to meet a child-count quota.
- Use
bd create, notbd q, when you need a real description. - Do not pass dependency flags to
bd create; create first, then wire withbd dep add. - Reconcile against the existing backlog before creating:
bd searchthe surface and title terms; attach, supersede, or reopen an existing bead before filing a new one. Audit-derived batches followth-engineering/subskills/greenfield-parity/references/idempotent-audit.md(ledger, audit-key, convergence record) so re-runs file nothing twice. - Stamp an execution-complexity label on every bead at creation:
complexity:low,complexity:medium,complexity:high, orcomplexity:epic. The coordinator maps this label straight to a worker model tier unless the design/specification override or Reconciliation Floor in../beads-coordinator/references/runtime-and-safety.mdapplies. Calibrate honestly:lowis docs/config/test-only or single-file mechanical work with no design judgment; when in doubt between two tiers, pick the lower — the coordinator escalates on evidence. - Keep non-lane work out of the dispatch queue with one of two labels, which
rank_ready_work.pyexcludes:no-dispatchfor tracking or meta beads, such as a trial log, a decision record, or an owner checklist.scope:local-statefor work on gitignored state that exists only in the main checkout (.beads/leftovers, caches, local databases), which no worktree reaches. Split a mixed bead so the tracked-file part stays ordinary lane work; the owner does the local-state part, or names a looper for it.
Read Order
| File | Read when | Content |
|---|---|---|
references/workflow.md |
Every real use of this skill | Research-first workflow, decomposition rules, creation sequence, and verification flow |
../../../th-projects/references/work-allocation.md |
Shaping implementation boundaries or dispatch packets | Proportional governing contracts, cohesive delivery boundaries, current packets, and readiness |
references/fields-and-examples.md |
Crafting issue text or choosing fields | Field schema, type/priority guidance, and example beads |
references/reconciliation-beads.md |
Writing or checking an epic | Finite outcome closeout, optional closeout-bead contract, and evidence prompt |
references/quality-checklist.md |
Before finalizing or after creating beads | Final checklist, persistence checks, and anti-patterns |
Trigger Sanity Check
- Should trigger: "Break this feature request into beads and make sure each child has enough context for a fresh worker."
- Should not trigger: "Implement bead
bd-123in its assigned worktree."
Resources
references/
workflow.md— Primary execution flow for research, decomposition, creation, and verification.fields-and-examples.md— Field schema, type selection guide, priority calibration table, and exemplary beads.reconciliation-beads.md— Finite outcome closeout and an optional verification prompt.quality-checklist.md— Final review checklist, persistence checks, and anti-patterns to remove.
04 同仓库更多 Skills
user-run
Use when a step needs privileges this session lacks (docker, sudo, another user's checkout or secrets, host access) and the owner must run it as tze. Writes a reviewable /home/orca/.tmp/<slug>/run.sh that impersonates tze (USER_TO_IMPERSONATE overrides), tees timestamped logs plus latest.log and collects artifacts there; Claude presents it with a risk and time brief, then continues from the logs. Triggers: "I'll run it for you", "permission denied on docker.sock", "needs sudo", "sudo: a password is required", "run as tze", "/user-run".
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
receiving-code-review
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation
finishing-a-development-branch
Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup
executing-plans
Use when you have a written implementation plan to execute in a separate session with review checkpoints
dispatching-parallel-agents
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies