askill
create-plan-rigorous

create-plan-rigorousSafety 90Repository

Create an extra-thorough implementation plan (sprints + atomic tasks) and get a subagent review. Use when the user wants a more rigorous plan than usual.

9 stars
1.2k downloads
Updated 3/8/2026

Package Files

Loading files...
SKILL.md

Create Plan (Rigorous)

Same as create-plan, but more rigorous: add per-task complexity notes, explicitly track dependencies, and get a subagent review before finalizing.

Contract

Prereqs:

  • User explicitly requests a more rigorous plan than normal.
  • You can spawn a review subagent.
  • plan-tooling available on PATH for linting (install via brew install nils-cli).

Inputs:

  • User request (goal, scope, constraints, success criteria).
  • Optional: repo context (files, architecture notes, existing patterns).

Outputs:

  • A new plan file saved to docs/plans/<slug>-plan.md.
  • A short response linking the plan path and summarizing key decisions/risks.

Exit codes:

  • N/A (conversation/workflow skill)

Failure modes:

  • Request remains underspecified and the user won’t confirm assumptions.
  • Plan requires repo access/info you can’t obtain.
  • Subagent review yields conflicting guidance; reconcile and document the decision.

Workflow

  1. Clarify (if needed)
  • If underspecified, ask 1–5 “need to know” questions first.
  • Use the format from $AGENT_HOME/skills/workflows/conversation/ask-questions-if-underspecified/SKILL.md.
  1. Research
  • Identify existing patterns and the minimal touch points.
  • Identify tricky edge cases, migrations, rollout, and compatibility constraints.
  1. Draft the plan (do not implement)
  • Same structure as create-plan, plus:
    • Fill a per-task Complexity score (1–10).
    • Explicitly list dependencies and parallelizable tasks.
    • Add a “Rollback plan” that is operationally plausible.
  1. Save the plan
  • Path: docs/plans/<slug>-plan.md (kebab-case, end with -plan.md).
  1. Lint the plan (format + executability)
  • Run: plan-tooling validate --file docs/plans/<slug>-plan.md
  • Fix until it passes (no placeholders in required fields; explicit validation commands; dependency IDs exist).
  1. Subagent review
  • Spawn a subagent to review the saved plan file.
  • Give it: the plan path + the original request + any constraints.
  • Explicitly instruct it: “Do not ask questions; only provide critique and improvements.”
  • Review rubric for the subagent:
    • Check for missing required task fields (Location, Description, Dependencies, Acceptance criteria, Validation).
    • Check for placeholder tokens left behind (<...>, TODO, TBD) in required fields.
    • Check task atomicity (single responsibility) and parallelization opportunities (dependency clarity, minimal file overlap).
    • Check that validation is runnable and matches acceptance criteria.
  • Incorporate useful feedback into the plan (keep changes minimal and coherent).
  1. Final gotchas pass
  • Ensure the plan has clear success criteria, validation commands, and risk mitigation.

Plan Template

Shared template (single source of truth):

  • skills/workflows/plan/_shared/assets/plan-template.md

Rigorous requirement:

  • Fill Complexity for every task (int 1–10).

Install

Download ZIP
Requires askill CLI v1.0+

AI Quality Score

82/100Analyzed 2/25/2026

High-quality skill with comprehensive workflow, clear contract, and structured subagent review process. Well-organized with good safety mechanisms (linting + review). Slightly penalized for internal-only path references and repo-specific assumptions, but the rigorous planning approach is valuable and reusable with adaptation. Tags (github-actions, linting) are somewhat mismatched with the skill content."

90
85
70
85
80

Metadata

Licenseunknown
Version-
Updated3/8/2026
Publishergraysurf

Tags

github-actionslinting