plan-review
Pre-implementation plan audit stress-testing scope, assumptions, risks, and failure modes before code is written. Triggers on: "review this plan", "is this plan solid", "what am I missing", "challenge my assumptions", "stress-test this", "/plan-review".
How do I install this agent skill?
npx skills add https://github.com/mathews-tom/armory --skill plan-reviewIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill is a read-only audit tool for technical plans. While it is safe in terms of system modifications, it processes untrusted user input without boundary delimiters, making it potentially susceptible to indirect prompt injection where instructions hidden in plans could influence the agent's review.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Plan Review Skill
Purpose
Execute a structured pre-implementation audit of a technical plan, proposal, or design document. The goal is to surface risks, bad assumptions, missing pieces, and scope problems before any code is written — when course corrections are cheapest.
This skill is read-only. It never modifies code. It produces a severity-tagged review document with a final ship/rethink/reject verdict.
Step 0 — Mode Selection
Use the requested lens: Product, Engineering, or Combined. When none is specified, use Combined without a configuration question. Resolve facts from the plan and repository; ask only for a material decision the evidence cannot settle.
Also assess scope size from the plan:
- Small change (single file, ~100 lines or fewer): deliver a compressed 4-section review — Scope, Risks, Missing, Verdict. Skip the full multi-section template.
- Standard change: execute the full review sequence below.
Full Review Sequence
1. Plan Comprehension
Read the plan and its authoritative project goal. Summarize the intended user workflow, the next useful outcome, non-goals and constraints in 2–3 sentences. Identify a mismatch between project value and technically coherent milestones before reviewing implementation detail. Do not pause for confirmation unless an unresolved contradiction changes the outcome or safety.
2. Assumption Challenge
Extract material assumptions that could change the next outcome or its safety. For each one:
| Assumption | If wrong? | Supporting evidence | What falsifies it? |
|---|
Common assumption categories to probe:
- Data availability and shape
- Third-party API stability and rate limits
- Team familiarity with chosen tools
- Performance characteristics of dependencies
- Backward compatibility requirements
- Deployment environment capabilities
3. Risk & Failure Mapping
For components on the next delivery path, identify demonstrated failure risks:
| Component | Failure Mode | Blast Radius | Recovery Strategy |
|---|
Additionally identify data flow shadow paths — side effects, async callbacks, event propagation, or cache invalidation chains that are not on the happy path but will execute in production.
Use ASCII diagrams to illustrate non-obvious data flow or failure propagation where the plan involves three or more interacting components.
4. Component-by-Component Review (Engineering Lens)
For each major component, assess:
- Error handling strategy — Are current failures explicit and recoverable where required? Do not prescribe a registry or retry framework without a consumer.
- Data integrity invariants — What invariants must hold? How are they enforced? What happens when they break?
- Concurrency and race conditions — Shared state, lock ordering, optimistic vs. pessimistic strategies, idempotency guarantees.
- Performance under load — Does the design meet evidenced workload and latency constraints? Hypothetical 10x capacity is not a gate without a source requirement.
- Test strategy adequacy — Do focused tests and an observed interaction cover the promised behavior and affected contracts? Separate fast local/PR feedback, integration-candidate checks and broad mainline/release coverage.
This section is language- and framework-agnostic. Reference
references/project-detection.md for framework-aware examples when the
user's stack is known.
Skip this section when running product-lens-only mode.
5. Scope & Priority Assessment (Product Lens)
- Necessary now vs. eventually useful — Does each milestone advance the next useful project outcome? For prerequisites and architectural expansion, name the consumer and concrete cost or risk of deferral.
- Deferral candidates — Defer optional capability with a revisit condition. Preserve correctness, security, data integrity and expensive-to-reverse constraints of the current capability.
- Changeability vs. speculative extensibility — Favor clear boundaries and affordable refactoring over unused frameworks, providers or deployment machinery.
- Planning horizon — Detail actionable milestones; keep distant outcomes provisional. Missing distant execution prompts are not defects.
- Delivery cost — Prefer one cohesive PR unless a split improves review or safe integration; examine stack depth and repeated CI/review waits.
Skip this section when running engineering-lens-only mode.
6. Integration Review
How components connect to each other and to the outside world:
- API contracts — Request/response shapes, versioning, error codes between modules.
- State management across boundaries — Who owns state? How is it synchronized? What happens during partial failure?
- Migration and deployment ordering — Which components must deploy first? Are there intermediate states where the system is inconsistent?
- Rollback compatibility — Can each deployment step be reversed independently? What data is irreversible?
7. What's Missing
Report a missing element only when a source requirement or concrete failing scenario makes it necessary for the current outcome. Consider security, data integrity, recovery, workload and operational needs at that boundary; do not turn this list into mandatory dashboards, runbooks, retries, benchmarks or extra qualification. Put optional improvements under non-blocking follow-ups, or omit them.
8. Execution Assessment
Evaluate the proposed implementation order:
- Dependency ordering — Are prerequisites built before dependents?
- Parallel work opportunities — Which tasks have no mutual dependency and can proceed simultaneously?
- Risk-first vs. value-first — Address an unknown early only when it threatens the next useful outcome or an expensive-to-reverse decision.
- Bounded experiments — Recommend a focused spike only when it can resolve that unknown more cheaply than implementation; state its decision and stop condition.
- Integration readiness — Keep required checks before merge against the actual integration candidate. Local checks and broad post-merge coverage complement rather than replace that safeguard.
9. Verdict
Deliver exactly one of:
| Verdict | Meaning |
|---|---|
| Ship | Plan is solid. Proceed as written. |
| Ship with changes | Viable, but specific modifications listed below are required before proceeding. |
| Rethink | Fundamental structural issues require re-planning. Itemize what must change. |
| Reject | Plan is not viable. Explain why and what alternative direction to consider. |
Include a one-paragraph rationale for the verdict.
Compressed Review (Small Changes)
For small-scope changes (single file, ~100 lines), deliver four sections only:
- Scope — What the change does and its boundaries.
- Risks — Failure modes and blast radius (brief table).
- Missing — Only material gaps required by this outcome.
- Verdict — Ship / Ship with changes / Rethink / Reject.
Interaction Protocol
- Batch related material decisions; finish the reachable review rather than pausing after each finding.
- Distinguish blocking defects from optional improvements. A plausible long-term benefit alone is not a blocker.
- This skill is read-only. It does not create, modify, or delete files.
- Use a diagram only when it clarifies a non-obvious dependency or failure.
Output Format
Return only applicable sections, severity-tagged findings with evidence and consequence, and a final verdict with rationale. Separate blocking changes from optional follow-ups. Omit empty sections and do not repeat findings in a second summary table. No finding is a valid result; do not invent work to fill a template.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/mathews-tom/armory/plan-review">View plan-review on skillZs</a>