$ loading_
Review implementation plans for gaps, assumptions, and sequencing before coding starts
This skill appears to be an open-source, prompt-only plan-review workflow with no required secrets, no declared remote endpoints, and no evident local code-execution capability, so overall risk is low. Minor attention is warranted because the README mentions reading/writing session memory and possible web validation, but the material reads like workflow guidance rather than a standalone high-privilege tool.
The material explicitly states that no keys or environment variables are required. There is no indication of API keys, tokens, or other sensitive credential collection, storage, or forwarding, so credential abuse exposure is low.
The metadata and objective checks indicate no remote host endpoints. Although the README mentions using web search to validate uncertain claims, it does not declare specific egress destinations, providers, or default data-upload behavior; based on the available material, no explicit network egress design is shown.
This is labeled as a prompt-only skill. The documentation describes plan review, parallel subagents, and synthesis workflow, with no installation scripts, shell commands, local process launching, or system-call permissions, so no concrete code-execution red flags are present.
The README refers to reading `/memories/session/plan.md` and saving output to `/memories/session/plan-review.md`, indicating expected access to and writing of session-level content. This is typical workflow data access rather than a clear over-permission red flag, but users should still consider whether the session contains sensitive information.
The source is an open-source GitHub repository under `microsoft/vscode-team-kit`, and source availability is a strong risk-reducing factor. However, the material also shows no declared license, 0 stars, and unknown maintenance status, so supply-chain confidence is generally positive but still warrants basic review.
Copy the install command and let the AI configure it · recommended for beginners
Please install the "review-plan" skill from askskill: 1. Download https://raw.githubusercontent.com/microsoft/vscode-team-kit/main/review-plan/skills/review-plan/SKILL.md 2. Save it as ~/.claude/skills/review-plan/SKILL.md 3. Reload skills and tell me it's ready
Please review the implementation plan below. Focus on missing requirements, incorrect assumptions, dependency risks, and whether the task breakdown and execution order are sound. Output your review as 'Issue, Impact, Recommendation' and conclude whether it is ready for implementation: [paste plan here]
A structured review highlighting plan gaps, risks, sequencing issues, actionable fixes, and a clear go/no-go conclusion.
Here is a plan.md produced by the Plan agent. Please do a pre-implementation review. Check whether the scope is clear, acceptance criteria are complete, and whether testing and rollback plans are missing. List required changes and optional improvements: [paste plan.md]
A delivery-focused review covering scope, acceptance, testing, rollback, and a list of required changes versus optional improvements.
Please review the step sequence in this development plan, identify ordering that may cause rework, blockers, or unnecessary risk, and reorganize it into a safer execution order. Explain the reason for each change: [paste plan here]
An improved execution sequence plus a step-by-step explanation of issues in the original order and why each change was made.
Your goal is to CRITICALLY review the given implementation plan. Provide thorough, constructive feedback that enhances the plan's quality and likelihood of successful execution and solving the stated goal.
Fan out parallel read-only subagents, each assigned a different plan-review area, then synthesize the highest-signal findings. Each subagent goes deep on one area, catching gaps a single pass misses.
Pick 2–4 areas based on the plan's complexity and risk profile.
| Area | Focus |
|---|---|
| Completeness | Missing requirements, unaddressed edge cases, gaps between the stated goal and proposed steps, unclear expected outcomes |
| Grounding | Technical soundness and factual accuracy — references to nonexistent or deprecated APIs, hallucinated functions, incorrect codebase assumptions, outdated versions. Use web search to validate uncertain claims. Suggest alternative approaches when the plan is fundamentally wrong |
| Sequencing | Dependency correctness between steps, parallelism opportunities missed, blocking-step identification, optimal ordering |
| Scope | Over-engineering, gold-plating, scope creep or under-specification, scope disproportionate to value delivered |
| Verification | Are verification steps specific, actionable, and covering the riskiest parts of the change? Missing test strategies |
| Risk | Unaddressed failure modes, migration risks, backward compatibility gaps, missing rollback strategy |
/memories/session/plan.md first.Launch 2–5 parallel subagents using the area prompts below. Each subagent works in isolation — do not share one area's findings with another before synthesis.
Each gets a self-contained prompt with its area, the plan location, and the return format. Subagents read the full plan from session memory themselves. Subagents may surface findings their area doesn't explicitly list — don't constrain them to only the listed focus items.
When all subagents return:
Always save the synthesized findings to session memory at /memories/session/plan-review.md. This makes them available for follow-up turns, plan revision, and cross-referencing during implementation.
/memories/session/plan.md), preserving the plan's existing structure.Before reporting a finding, ask: Does this change what a developer would actually do — or how likely the plan is to succeed? Ground every finding in something observable: a file that doesn't exist, an API that behaves differently, a requirement the steps don't cover, a sequence that creates rework.
…
Add emoji reactions to GitHub issues or pull requests quickly.
Let AI agents read and write memory with environment-aware storage fallback.
Gather independent multi-model plans and debates for implementation and architecture decisions.
Get high-signal second opinions on plans, designs, and implementations early.
Create and manage AST ban rules to block specific code syntax patterns.
Set up Component Explorer with CLI, MCP, and VS Code tooling.
Run multi-model reviews for code changes, PRs, and risky edits.
Execute detailed plans in batches with review checkpoints and course correction.
Create step-by-step implementation plans for engineers new to a codebase.
Review code deeply across correctness, tests, security, performance, and product quality.
Create detailed implementation plans for engineers with little or no codebase context.
Create objective-driven plans for agent-led development with an audit trail.