Sdd draft tasks
Skill andresnator/agents-orchestrator/skills/sdd-draft-tasks
Trigger: draft tasks, borrador de tareas, SDD tasks interview. Draft ordered, verifiable tasks from spec and design; plan-only, write on approval.From its SKILL.md
npx -y skills add andresnator/agents-orchestrator --skill sdd-draft-tasksAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
4.0 KB, 820 tokens by cl100k_base, as published. Nobody here has run it
Activation Contract
Use when drafting an OpenSpec tasks.md from an approved spec and design: an ordered, dependency-aware checklist. Works standalone or delegated by grill.
Hard Rules
- Follow the
grillingskill: one question at a time, recommendation attached, stop and wait; read spec/design context instead of asking when possible. - Follow the
native-question-uxskill for question presentation; ask one question at a time and stop after each answer. - Interview/summaries/gates use the user's language; artifacts default to English unless Spanish artifacts are explicitly requested.
- Plan-only: read-only codebase access; no code edits, builds, installs, tests, or state changes. Only write planning
.mdfiles after explicit approval. - Every task line MUST be
- [ ] X.Y {concrete action naming real files}: Specific, Actionable, Verifiable, Small enough for one session. Never vague ("implement feature"). Testing tasks reference specific spec scenarios. - Order groups by dependency: a task may only depend on earlier tasks; never forward-reference a later group. Execution scheduling (batching, parallelism, worktrees) is left to the implementer; the
Files:scopes exist so it can schedule safely. - Every group carries a
Files:line right under its heading: the directories or globs its tasks will touch. Scopes are scheduling predictions, not enforcement — only groups with disjoint scopes and no shared hotspot may run as parallel waves downstream; a missingFiles:line means that group serializes. - Fill the
Shared hotspots:guard line with repo-wide collision points any group might touch (lockfiles/manifests, barrel or index files, DI/route registries, generated code, shared fixtures), ornone. - Preserve the five forecast guard lines VERBATIM (the orchestraitor gates and schedules on them):
Decision needed before apply:,Chained PRs recommended:,Chain strategy:,400-line budget risk:,Shared hotspots:. - Keep the artifact under 650 words. When delegated by grill, return the approved draft and do not write files; the orchestrator owns the single write step.
Decision Gates
| Situation | Action |
|---|---|
| Standalone and spec/design exist on disk | Read them; the design File Changes table seeds concrete task targets. |
| Missing spec or design | Offer to run sdd-draft-spec or sdd-draft-design first. |
| A task depends on another | Order it after its dependency; never forward-reference a later group. |
| Estimated changed lines exceed 400 | Recommend a chained-PR split in the forecast and ask. |
Execution Steps
- Read
assets/tasks-template.mdandreferences/question-bank.md; read the approved spec and design. - Interview for slicing, dependency order, and per-task verification.
- Draft
tasks.mdinline from the template with the forecast and dependency-ordered grouped checkboxes. - Present for approval; revise until approved.
- If standalone and approved, ask before writing
.ai/orchestrator/changes/{change-name}/tasks.md.
Output Contract
When the caller owns the write (standalone or delegated by grill): return approved/unapproved status, tasks draft, dependency order, forecast values, and open questions. When loaded by a phase agent that writes the artifact itself: the drafting rules and template above apply, and the return is the caller's own output contract — never the full artifact. If a referenced skill cannot be resolved by name in the current runtime, say so instead of silently continuing.
References
assets/tasks-template.mdreferences/question-bank.mdgrillingskillnative-question-uxskill
Attribution
Inspired by Matt Pocock's grilling skills at https://github.com/mattpocock/skills; adapted for OpenSpec task drafting.
What ships with it: 2 files
2.1 KB alongside SKILL.md
assets/
- tasks-template.md1.4 KB
references/
- question-bank.md761 B