create-fred
Skill1 starsChanged 25 days agoApache-2.0View on GitHub
Create or refine a Feature Requirement Document (FRED) in the current repository. Use to turn a feature idea into a scoped specification, gather requirements, resolve implementation questions, or make a Draft ready for implementation. Produces a FRED plus roadmap and delivery-task updates.
In this repository2 other files it gives its agents.
- fredSkill
- implement-fredSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Follow applicable AGENTS instructions and the project's agent entrypoint, normally docs/agent-guides/00-AGENT-ENTRYPOINT.md.
- Distinctive: line 24
Use the project's legacy index/layout if it has not migrated.
- Very specific: line 26
If FRED infrastructure is missing, use the available fred bootstrap workflow when requested or accepted.
- Distinctive: line 53
Do not invent answers, invent a separate decision lifecycle stage, or treat silence as approval.
- Distinctive: line 57
When the runtime can spawn subagents, use multiple in parallel for complex specification work: researching distinct contracts, inspecting …
- Distinctive: line 69
Use the project template; allocate the next unused stable ID from ROADMAP (or its legacy equivalent).
- Very specific: line 74
Add schema/API/event, UI, migration, timeout/recovery, or policy detail only where it changes implementation decisions.
- Distinctive: line 87
Create applicable OPERATIONS items now for external setup and shared deployed journeys.
- Distinctive: line 95
Write or update the FRED and synchronize ROADMAP and applicable OPERATIONS links.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Create FRED
Turn the user's intended outcome into a slice another agent can implement without guessing. Work conversationally; ask only questions whose answers materially affect scope, behavior, contracts, or verification.
Read the project contract
Follow applicable AGENTS instructions and the project's agent entrypoint,
normally docs/agent-guides/00-AGENT-ENTRYPOINT.md. Read its FRED template,
ROADMAP, existing related slices, and relevant architecture/code. Load procedure
guides only when needed. Project documents and user instructions govern scope,
paths, lifecycle, commands, and permissions; this skill adds a creation workflow.
Use the project's legacy index/layout if it has not migrated. Do not load a
sibling skill by relative filesystem path: this skill can be installed alone.
If FRED infrastructure is missing, use the available fred bootstrap workflow
when requested or accepted. Otherwise continue useful requirement gathering
and offer bootstrap before creating a competing document structure. Do not
silently migrate an existing project or code the feature during specification.
Gather and resolve requirements
- Establish the intended user/operator, outcome, current behavior, and scope. Reuse supplied context and existing requirements rather than asking again.
- Inspect affected code and architecture for actual capabilities, reusable contracts, and constraints. Separate implemented facts from proposals.
- Identify consequential unknowns: permissions, data/state ownership, external behavior, compatibility, error recovery, and deployment/configuration needs. For each blocking open decision, give context and suggestions (a recommendation plus tradeoffs), then require explicit user acknowledgment. Progress on independent inspection while answers are pending.
- Research genuinely uncertain contracts or integrations using the installed versions and relevant official sources. Record source links and assumptions; browsing documentation is not proof of a live integration. Avoid research unrelated to a decision the slice needs.
- Propose an implementable slice with explicit non-goals. If the requested outcome needs several slices, explain the split and integration points; respect the user's scope before expanding into a product-wide backlog.
Straightforward requests need no mandatory interview. Use stated assumptions for low-impact choices; unresolved decisions that materially change the implementation keep the FRED Draft with the open questions recorded there. Do not invent answers, invent a separate decision lifecycle stage, or treat silence as approval. Draft → Open requires explicit user acknowledgment of every blocking open decision.
When the runtime can spawn subagents, use multiple in parallel for complex specification work: researching distinct contracts, inspecting separate code areas, or drafting independent FRED sections. Synthesize results in the coordinating session. Keep ID allocation, conflicting shared-surface decisions, and Draft → Open readiness sequential. When the runtime allows choosing a model for subagents, prefer a lighter/faster model for bounded parallel work (research, section drafts, scoped exploration) unless the slice needs stronger reasoning; keep the coordinating session on the stronger model if that is how the user invoked the skill.
Specify the slice and its place in the roadmap
- Use the project template; allocate the next unused stable ID from ROADMAP (or its legacy equivalent). Account for all existing states/history and concurrent allocation; never reuse an ID. Preserve the ID when refining a draft.
- Describe goal/non-goals, observable behavior, contracts/affected surfaces, tasks, implementation acceptance, evidence expectations, and delivery links. Add schema/API/event, UI, migration, timeout/recovery, or policy detail only where it changes implementation decisions.
- Assign a bounded-outcome track and identify shared contract/migration owners. Distinguish prerequisites requiring an agreed interface from those requiring implemented code. Allowed fakes name assumptions, limitations, and the owner of real integration. Check for dependency cycles and shared-surface conflicts.
- Define proportionate implementation checks on the FRED. Include browser or render inspection only when verification criteria call for UI/browser QA; backend slices retain relevant rule, authorization, persistence, and integration checks when those matter. Include the project's global lint/format check (from the entrypoint) before implementation review handoff and Implemented/Closed sign-off, including cascading monorepo checks; module checks cannot replace it.
- Create applicable OPERATIONS items now for external setup and shared deployed journeys. Each has an owner, prerequisites, what it blocks, target, executable steps, pass condition, and evidence destination. One flow may cover several FREDs. Setup needed to discover a contract or code behavior is also an implementation prerequisite, not merely a closure gate.
Write and hand off
Write or update the FRED and synchronize ROADMAP and applicable OPERATIONS links. Keep paths stable. Mark Open only when scope/contracts and local acceptance are sufficiently defined under project rules and every blocking open decision has explicit user acknowledgment; otherwise retain Draft with the exact unresolved question, suggested options, and next action. An Open slice may still be blocked by an explicit prerequisite. Planning-only Draft → Open does not require running code checks or claim implementation/verification success.
Link existing architecture and guides. Record relevant research where it is useful, and label proposed architectural changes; do not describe them as implemented. Add a new document only for knowledge that merits maintenance. Record justified changes to existing acceptance or verification expectations, without weakening them merely to permit sign-off. Keep secrets out of specs.
Finish with the FRED link, state/track, consequential assumptions or unresolved questions, and implementation readiness. When specification stops unfinished, leave concise resume notes. Honor existing Git/deploy authorization; when committing, follow the entrypoint Git/PR policy. Allocating a slice or creating an operator task does not authorize external actions.
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.

