sdd-orchestrate-prds
Skill3 starsChanged 36 days agoMITView on GitHub
Slice SDD when a broad request must become several cohesive PRDs under a shared prefix; for a single PRD, use sdd-create-prd.
In this repository26 other files it gives its agents.
- AGENTS.mdAGENTS.md
- ai-ready-fixSkill
- ai-ready-scoreSkill
- architectural-analysisSkill
- commitSkill
- deep-reviewSkill
- domain-modelingSkill
- dotnet-efficient-validationSkill
- herdr-orchestrationSkill
- no-workaroundsSkill
- repository-cli-efficiencySkill
- sdd-create-prdSkill
- sdd-create-techspecSkill
- sdd-execute-correctionsSkill
- sdd-execute-taskSkill
- sdd-orchestrate-flowSkill
- sdd-orchestrate-tasksSkill
- sdd-plan-auditSkill
- sdd-plan-correctionsSkill
- sdd-plan-refactoringSkill
- sdd-plan-tasksSkill
- sdd-review-codeSkill
- sdd-snapshotSkill
- sdd-triageSkill
- update-nuget-packagesSkill
- writing-agents-mdSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
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
- Distinctive: line 15
Output: every obligation from step 1 belongs to exactly one slice.
- Very specific: line 20
After each slice except the last, run the session pause from session-continuity.md with stage planning and the next slice as next step.
- Distinctive: line 25
Read each source once per version.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Orchestrate SDD PRDs
A broad request becomes a set of slices: each slice has one primary outcome, its own PRD, and a TechSpec that fits without branching. The user approves the slicing before any PRD exists on disk. This session writes every PRD; subagents are read-only explorers.
- Understand the request and fix the prefix. Gather objective, users, journeys, surfaces, integrations, constraints, and declared out-of-scope items. Consult local evidence before public sources and mark each obligation as fact, assumption, or product decision. Derive the prefix in kebab-case from the request's domain, or use
--prefixwhen supplied. List existingtasks/prd-*-[prefix]-*folders and reuse the matching slug when the scope coincides; a completed feature's folder is not reopened, and the new scope becomes a new slice at the next free number. If one of them holds anactivecontext-snapshot.md, load it through the Load branch of.agents/skills/sdd-snapshot/SKILL.mdand resume from its next step; a slicing it records as approved is not asked again unless the request changed. Output: prefix resolved, obligations inventoried with their origin, and already-written slices identified; ask only for the missing information that blocks the inventory. - Design the slicing. Group the obligations by cohesion: one primary outcome per slice, minimal coupling between slices, and a scope one TechSpec covers within a single subsystem. Separate independent journeys, distinct surfaces, and layers deliverable in different orders; keep together whatever shares a contract and would fail alone. Assign each cross-cutting obligation (configuration, observability, NFR) to one owning slice and cite it as a constraint in the others. Number the slices in dependency order, continuing from the highest number already used under
tasks/, and name each slug[NN]-[prefix]-[slice]. Output: every obligation from step 1 belongs to exactly one slice; each slice has a single primary outcome; dependencies between slices form an order without cycles. A single slice is a valid result when the request has one primary outcome. - Approve the slicing. Present each slice with slug, outcome, covered obligations, and dependencies. Use the available question tool (
AskUserQuestion) or, without it, a textual question carrying the same options, each naming the slugs it produces: the proposed slicing, a more granular alternative, and a more grouped one. Rely on the automatic "Other" option for the user to type whatever split they want, including a different prefix. Remap the obligations onto the chosen split and ask again only when it leaves an obligation without an owner or creates a cycle. With more than one slice, write the snapshot withsdd-snapshotin the first slice's folder right after approval, recording the approved slugs, outcomes, owned obligations, and the user's answer asnowdecisions, so a later session end cannot lose the slicing. Output: slicing approved in the conversation, remapped onto the inventory, and recorded in the snapshot; no PRD on disk until that answer. - Draft each slice. For each approved slice, in dependency order, follow
.agents/skills/sdd-create-prd/SKILL.mdin this session with the final slug, that slice's obligations and evidence, the sibling slugs as boundaries, and the instruction to record cross-slice dependencies underConstraints and dependenciesand boundaries underOut of scope. Load only that slice's obligations and cited sources before drafting it. Read-only explorers may gather evidence for several slices in parallel; this session writes every PRD. After each slice except the last, run the session pause fromsession-continuity.mdwith stageplanningand the next slice as next step; a written snapshot moves to the next slice's folder and the previous one is markedsuperseded. Output: each approved slice hastasks/prd-[NN]-[prefix]-[slice]/prd.mdwritten, or a blocker with a concrete cause that stops only that slice. - Check the set and report. Check coverage against step 1's inventory, unique IDs within each PRD, each functional requirement appearing in a single slice, and dependencies declared on both sides. Correct a divergence in the owning slice's PRD before reporting. Report paths, the suggested order for
sdd-create-techspec, and pending items inherited from the PRDs. Mark the snapshotclosed, or point its next step at the first slice's TechSpec when the user continues. Output: reviewable set on disk; obligations without an owner and repeated requirements resolved or explicitly pending.
Read each source once per version. When a request extends an existing prefix, preserve the written PRDs and treat the new scope as additional slices starting from the next free number. Under sdd-orchestrate-flow, the flow runs the session pause and records the slicing in workflow.md.
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
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.

