agentsclimarketplace

Plan generator

Skill hekivo/superpowers-sage/skills/plan-generator

Generate implementation plans from approved designs — plan-generator, phase-based plans, parallel batch execution, task dependency graph, owner skill routing, acceptance criteria, global done criteria, plan frontmatter, layout contract, interaction contract, anti-drift rules, plan format, AD-2 byte-for-byte gate, scope definition, execution plan, phases and batches, task anatomy, plan file formatFrom its SKILL.md

Install
npx -y skills add hekivo/superpowers-sage --skill plan-generator

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 13 stars13 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.

SKILL.md

3.6 KB, as published. Nobody here has run it

Plan Generator

Generate implementation plans from approved architecture specs.

Announce at start: "I'm using plan-generator to transform the approved architecture into an executable implementation plan."

Input

$ARGUMENTS

HARD GATES

  • Do not generate a plan without an approved architecture spec.
  • Do not implement code in this skill.
  • Do not mark the plan complete if dependencies are ambiguous.

Procedure

1) Resolve and validate spec input

Accept one of:

  • Direct spec path from arguments
  • Most recent approved file in docs/superpowers/specs/

Reject if spec approval is missing.

2) Preflight — handoff payload validation

Before generating any plan files, validate the architecture spec against its claimed sources (git refs, external APIs, content models). See references/anti-drift.md for the full preflight protocol.

If the spec's "chosen approach" is "zero-migration port from legacy schema", emit an AD-2 byte-for-byte gate per ported component. See references/anti-drift.md.

3) Parse architecture into implementation units

Extract:

  • Components and responsibilities
  • Data model decisions
  • Integration points
  • Quality constraints

Translate each into concrete implementation units.

4) Build dependency graph

Classify every unit:

  • parallel: independent work
  • sequential: depends on prior unit output
  • gated: requires user/reviewer checkpoint

Document graph in the plan.

5) Assign execution strategy and skill routing

For each unit, define:

  • Strategy: autonomous, interactive, or mixed
  • Suggested skill(s): superpowers-sage:building, superpowers-sage:modeling, superpowers-sage:designing, superpowers-sage:verifying, and reference skills (acorn-*, wp-*) as needed
  • Acceptance criteria

6) Generate plan directory and files

Create the directory and file structure as defined in references/plan-format.md.

Full file templates and frontmatter schemas: skills/plan-generator/references/plan-format.md

7) Validate plan consistency

Check:

  • Every component appears exactly once
  • Every dependency has a source node
  • No component is both parallel and sequential in the same stage
  • Strategy aligns with risk/complexity
  • Every interactive component has a visual checkpoint owner

Full validation rules: skills/plan-generator/references/anti-drift.md

8) Final handoff for execution

Present concise execution summary (phases, parallel batches, checkpoints) and offer superpowers-sage:building.

Verification

Before completion, confirm all items:

  • Approved source spec exists and is referenced in source-spec
  • Plan directory exists with required files
  • Dependency graph has no unresolved nodes
  • Each component has acceptance criteria in its sub-plan

Failure modes

  • Spec missing or unapproved: block and request architecture approval.
  • Dependency cycle detected: surface cycle and propose reordering.
  • Incomplete component definitions: return to architecture spec for clarification.
  • Too many cross-component dependencies: split into phased plan and mark phase gates.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.