AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Create Plan

skill-itsjavi-skills-create-plan · by itsjavi

A Claude skill from itsjavi/skills.

No reviews yet
0 installs
30 views
0.0% view→install

Install

$ agentstack add skill-itsjavi-skills-create-plan

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-itsjavi-skills-create-plan)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
4mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Create Plan? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Plan Creator

Create or update an implementation plan inside an already established planning workflow.

Do not use this skill to bootstrap, upgrade, or repair the workflow itself. If the project has no standalone /GUIDE.md, or the user asks to modernize the workflow, use setup-specs-workflow instead.

Workflow

  1. Resolve the target planning root:
  • Use the root named by the user when provided.
  • Otherwise prefer a directory with GUIDE.md and MILESTONES.md.
  • Check common roots in this order: .specs/, docs/, project-specs/, .agents/specs/, ai/.
  • If multiple plausible roots exist, choose the one whose GUIDE.md describes the active workflow or ask when the

answer is not clear.

  1. Read project context in tiers:
  • Always read only the root agent guide if present, /GUIDE.md, PRODUCT.md, MILESTONES.md,

BUG_FIXES.md, BUSINESS_RULES.md, COORDINATION.md, CHECKS.md, and MANUAL_QA.md.

  • Read the active or requested milestone record before opening related plans or checkpoints.
  • Read DESIGN.md, decisions, business-rule records, bug-fix records, research notes, setup docs, security docs, env

docs, and implementation files only when they affect the requested plan.

  • Use rg, filename scans, and index rows to find relevant records before opening full files.
  • Do not bulk-read all plans, checkpoints, decisions, or business-rule records.
  1. Minimum context sources:
  • root AGENTS.md, CLAUDE.md, or .cursor/rules if present
  • /GUIDE.md
  • /PRODUCT.md
  • /MILESTONES.md
  • /BUG_FIXES.md
  • /BUSINESS_RULES.md
  • /COORDINATION.md
  • /CHECKS.md
  • /MANUAL_QA.md
  1. Treat /GUIDE.md as the workflow source of truth for status vocabulary, numbering, file naming,

template usage, spec revision rules, checkpoint expectations, and index ownership.

  1. Check whether the request belongs in a bug-fix record instead of a plan:
  • Use /templates/BUG_FIX.md and update /BUG_FIXES.md when the work is primarily a

scoped defect report and fix proposal that does not need milestone sequencing, multiple implementation phases, or roadmap visibility.

  • Continue with plan creation when the fix expands into broad feature work, durable architecture changes, migrations,

cross-domain behavior, or multi-phase delivery.

  1. Determine the target milestone:
  • Use the milestone named by the user when provided.
  • Otherwise use the active milestone from /MILESTONES.md.
  • If there is no active milestone, pick the clearly requested milestone or ask whether to create/select one.
  1. Choose PPP for a new plan:
  • Use the user's requested plan number only when it is exactly three digits and unused inside the target milestone.
  • Otherwise pick the next unused three-digit plan number after scanning /plans/ and the target

milestone record's Drafted Plans section.

  • Never reuse cancelled, retired, or superseded plan numbers.
  1. Create or update /plans/MMM-PPP-kebab-case-title.md:
  • Prefer /templates/PLAN.md when present.
  • If the template is missing, follow the plan requirements in /GUIDE.md.
  • Draft phases A, B, C... with the final phase named Final review pass.
  • Include definition of done, out of scope, phase acceptance checks, validation plan, manual QA impact, circuit

breakers, risk notes, changelog impact, documentation updates, and decision-record check.

  1. Update the target milestone record:
  • Add or update the plan row in the milestone record's Drafted Plans section.
  • Keep plan registries and phase maps in the milestone record, not in MILESTONES.md.
  1. Update /MILESTONES.md only when cross-milestone focus changed: active milestone, recommended next

plan, latest checkpoint, milestone status, or recommended execution order.

  1. Apply the spec revision rule: material changes to accepted or active plans require an explicit plan update before

implementation continues, and unplanned work that changes behavior, architecture, configuration, APIs, operational flows, security posture, manual QA coverage, verification commands, or user-facing workflows must update the closest relevant spec in the same turn.

  1. Preserve git index state. Do not stage, unstage, commit, amend, reset, or discard files unless explicitly asked.

Plan Content Requirements

Include:

  • status table with all phases initially 🧭 Proposed
  • goal
  • definition of done with testable outcomes
  • out of scope
  • phases with goal, likely files or areas, deliverables, and acceptance checks
  • test / validation plan that references /CHECKS.md
  • manual QA impact that references /MANUAL_QA.md
  • changelog impact using Added, Changed, Fixed, Removed, Security, Operations, QA / Verification, or

None, plus a short human-readable note when release-visible

  • circuit breakers and stop conditions
  • risk notes
  • decision-record check with links to created, existing, or superseded decisions
  • documentation update checklist
  • final Final review pass phase

Circuit Breakers

Every plan should tell implementation agents when to stop instead of looping:

  • Stop after two repeated failures of the same check with no new evidence or changed approach.
  • Stop when requirements, business rules, decisions, or implementation constraints conflict.
  • Stop when required verification cannot run and no documented fallback exists.
  • Stop when the work expands beyond the active plan's scope.
  • Stop before changing public behavior, security posture, data model, deployment flow, or manual QA coverage unless the

relevant spec update is included.

Implementation Offer

After creating or updating the plan and reporting the files changed, ask the user whether they want implementation to start now.

Prefer a native choice UI when the host makes one available:

  • In Codex, if a request_user_input style tool is available, use it before ending the turn. Ask: "Implement this plan

now?" with choices:

  • Implement this plan - start implementing from the new plan in this thread.
  • Just save the plan - stop after the plan handoff.
  • In Cursor or another host with an equivalent native quick-pick/choice UI, use the closest equivalent.

If no native choice UI is available, end the final response with a concise plain-text question:

Want me to start implementing this plan now?
- Implement this plan
- Just save the plan

Do not begin implementation until the user chooses or clearly says yes. If the user chooses implementation, continue from the plan using the project's normal plan/checkpoint workflow. If a separate fresh-agent handoff is more appropriate, offer or generate an implementation prompt instead of editing code immediately.

Final Response

Report:

  • plan or bug-fix record created or updated
  • milestone record or BUG_FIXES.md row added or changed
  • plan or bug-fix number and status used
  • related checks/manual QA/spec docs updated or left as TBD
  • checks run, if any

Keep the response concise. Do not paste the full plan or bug-fix record unless the user asks.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.