workflow-feature-spec
Skill0 starsChanged 3 months agoMITView on GitHub
Run the full feature-spec pipeline for a new product / feature with commercial stakes — biz-brief (business context) → superpowers:brainstorming (technical spec) → superpowers:writing-plans (implementation plan). Use when the user says "let's build X", "I want to ship Y", "spec out this feature", "go from idea to plan for Z", or when starting any new feature with a user / market / business model. Do NOT use for purely internal tooling (skip biz-brief; go straight to brainstorming) or for bug fixes (use debugging-and-error-recovery instead).
In this repository11 other files it gives its agents.
- adversarial-diff-verifySkill
- agent-harnessSkill
- agents-md-generatorSkill
- plugin-manifest-referenceSkill
- publish-postSkill
- research-sweepSkill
- workflow-designSkill
- workflow-monthly-skill-auditSkill
- workflow-stage-draftSkill
- workflow-vault-weeklySkill
- work-trackingSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 51
Default: run brief → spec → plan in order, with checkpoints between each.
- Very specific: line 55
If <step> is not in {brief, spec, plan}, print:
- Distinctive: line 64
Before running the resume step, confirm the listed inputs exist.
- Very specific: line 81
Check for docs/superpowers/STATE.md.
- Very specific: line 91
After the pipeline completes, append to docs/superpowers/STATE.md:
- Very specific: line 96
If a stage consistently required retries across multiple runs, update this SKILL.md's stage description with the failure mode.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Skill: /workflow-feature-spec
Pipeline that takes a fuzzy feature idea and ends with an executable implementation plan grounded in business intent + technical spec.
When to use
- New feature / product / initiative with commercial stakes.
- The user is at the "let's build X" stage; no spec yet.
When NOT to use
- Internal tooling with no user/market layer → call
superpowers:brainstormingdirectly. - Bug fix → use the debug-and-fix flow instead.
- Already have a spec → call
superpowers:writing-plansdirectly.
Steps
| Step name | Stage | Invokes |
brief | Business context | biz-brief |
spec | Technical spec | superpowers:brainstorming |
plan | Implementation plan | superpowers:writing-plans |
Stage 1: Business context — step brief — invokes biz-brief
Generates a 1-page Business Context Brief and saves it to docs/superpowers/briefs/YYYY-MM-DD-<feature>.md.
Output: brief file exists. Checkpoint: confirm brief is accurate before brainstorming.
Stage 2: Technical spec — step spec — invokes superpowers:brainstorming
Brainstorms the technical spec with the brief loaded as context. Saves spec to docs/superpowers/specs/YYYY-MM-DD-<feature>.md.
Output: spec file exists. Checkpoint: confirm spec coverage before writing the plan.
Stage 3: Implementation plan — step plan — invokes superpowers:writing-plans
Reads the spec and produces a task-by-task plan at docs/superpowers/plans/YYYY-MM-DD-<feature>.md.
Output: plan file exists, ready to hand off to subagent-driven-development or executing-plans.
Resume protocol
Default: run brief → spec → plan in order, with checkpoints between each.
--resume-from <step>: skip every step before <step> and begin from <step>. The named step itself runs.
If <step> is not in {brief, spec, plan}, print:
Invalid resume-from step: <step>
Valid steps: brief, spec, plan…and stop.
Required state when resuming
Before running the resume step, confirm the listed inputs exist. If anything is missing, prompt the user for it.
--resume-from brief— same as a fresh run; needs only the feature name.--resume-from spec— needs the brief atdocs/superpowers/briefs/YYYY-MM-DD-<feature>.md. If absent, ask the user to paste it or run frombrief. Brainstorming loads the brief as context.--resume-from plan— needs the spec atdocs/superpowers/specs/YYYY-MM-DD-<feature>.md. If absent, ask the user to paste it or run fromspec. Writing-plans operates on the spec.
Examples:
/workflow-feature-spec --resume-from spec # brief already exists, brainstorm from it
/workflow-feature-spec --resume-from plan # spec is approved, just write the planOther flags
--no-pauseto run end-to-end (rare — usually you want the checkpoints here).
Session open
Check for docs/superpowers/STATE.md. If it exists, read Verified facts, General rules, and Last session before starting — these carry forward what the last run learned so this run doesn't re-derive it.
Stage quality gate
After each stage completes, before the checkpoint: does the output actually satisfy the stage goal (brief = business intent clear; spec = technical approach decided; plan = tasks are executable)? If it partially misses, retry the stage once with an explicit note on the gap. On second failure, surface the issue to the user rather than continuing.
Stage validation should use a separate verification pass (re-read the output with fresh eyes as a critic, not as the author) rather than the author confirming their own work.
Session close
After the pipeline completes, append to docs/superpowers/STATE.md:
- Any decisions that were non-obvious and should not be re-derived next time
- Any stage that required a retry and why
- The resume pointer (last step completed, next step if interrupted)
If a stage consistently required retries across multiple runs, update this SKILL.md's stage description with the failure mode.
Failure handling
Brainstorming and writing-plans both have their own surface-assumptions / push-back behaviors built in. The workflow does not paper over those — when a sub-skill asks a clarifying question, the workflow surfaces it to the user. If a stage errors, stop and print the failed step plus the resume command (/workflow-feature-spec --resume-from <step>).
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.

