agentleFS
mab-gomab-go/xmind-mcp/.agents/skills/review-plan/SKILL.md

review-plan

Skill9 starsChanged 26 days agoMITView on GitHub

Reviews implementation plans, including any associated task items, for accuracy, correctness, clarity, and completeness. Use when invoked via /review-plan, when the user asks for a plan review, before implementing a plan or handing it to another agent, or as a self-check after producing or updating a plan.

37 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use implement-plan, review-design and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • JasonPaffSkill

    implement-plan

    Execute implementation plan with structured tracking and validation using subagent architecture

    Also in 5 projects that use this

  • dirkkok101Skill

    review-design

    Use before /plan, after /technical-design, when user says "review design", "check the design", or when design quality is uncertain.

    Also in 3 projects that use this

  • SvampenSkill

    review-implementation

    Reviews implementation against session/work stream goals before committing. Identifies shortcuts, hacks, incomplete work, and validates accomplishments match actual changes. Use before /session-end to ensure implementation quality.

    Also in 3 projects that use this

  • NVIDIASkill

    review-code

    Review a code repository for concrete issues in correctness, robustness, security, maintainability, tests, and integration. Use for repository-wide or focused code review where practical engineering judgment and strict scope control matter.

    Also in 7 projects that use this

  • PydanticSkill

    review-security

    Security review of the current branch against its merge base — sandbox escapes, memory errors, panics and resource-limit bypasses. Use when reviewing changes for security risk, or before merging anything touching the heap/ module, path_security.rs, the wire protocol or the pool.

    Also in 3 projects that use this

  • am-willSkill

    planner

    Create comprehensive, phased implementation plans with sprints and atomic tasks. Use when user says: "make a plan", "create a plan", "plan this out", "plan the implementation", "help me plan", "design a plan", "draft a plan", "write a plan", "outline the steps", "break this down into tasks", "what's the plan for", or any similar planning request. Also triggers on explicit "/planner" or "/plan" commands.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 7 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the project's agent instructions (e.g.

    line 8
  2. Very specific:

    File paths: Run ls or find to confirm referenced files exist.

    line 20
  3. Distinctive:

    API behavior claims: Read the relevant source to verify.

    line 22
  4. Very specific:

    Dependency claims: Check the dependency manifest (e.g.

    line 23
  5. Distinctive:

    Invariants: If the project's agent instructions declare invariants (things that must never happen), any step that would break one is a …

    line 45
  6. Distinctive:

    If the plan has associated task items (YAML frontmatter todos, a checklist, or structured tasks), review each item for:

    line 74
  7. Distinctive:

    Ambiguous: No clear completion criterion.

    line 78
  8. Distinctive:

    If required context is missing, ask targeted clarifying questions before proposing fixes.

    line 99
  9. Distinctive:

    If the plan is solid, say so briefly.

    line 100
  10. Very specific:

    After plan approval and implementation, run /verify-changes to confirm the build passes.

    line 108

“Common” is measured: kept by at least 6 in 10 of the 36 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Plan Review

Read the project's agent instructions (e.g. AGENTS.md), plus any nested AGENTS.md in directories the plan touches, before applying this skill.

Review directive

The agent implementing this plan may not be you -- it may be an agent without your context. The plan must be rock-solid: accurate, unambiguous, and complete enough to execute without guessing. But do not over-explain for its own sake. Assume a capable implementer.

Verification procedures

For each category, here is how to verify -- not just what to check.

Accuracy

  • File paths: Run ls or find to confirm referenced files exist.
  • Names: Confirm every function, type, config key, and build or script target the plan names exists and has the expected signature or definition. Use LSP (gopls) for Go symbols and grep for everything else.
  • API behavior claims: Read the relevant source to verify.
  • Dependency claims: Check the dependency manifest (e.g. go.mod, package.json).

Correctness

  • Execution simulation: For each step, mentally simulate: does the preceding step produce what this step needs as input?
  • Contradictions: Does step N undo or conflict with step M?
  • Broken intermediate states: Would the plan leave the project broken if stopped after any given step? Common cases:
    • Something defined but never registered or wired in where the project discovers it
    • A new entry of a kind the project lists in several places (registries, docs, config) that misses one of them
    • A check, test, or tool that hard-codes a list the change extends, so it silently skips the new entry
    • A new dependency direction that the project's architecture rules or linters forbid
    • New permissions or credentials that lack access to something they need indirectly
    • Generated artifacts not regenerated after their source changes
    • In this project, also check for:
      • A new tool not registered in internal/server/server.go, or missing from expectedAnnotations in internal/server/tools_test.go or contractCases in internal/server/error_contract_test.go
      • A new JSON key on a tree type missing from that type's field table in internal/xmind/json_codec.go, so it is kept as opaque data instead of populating the typed field; or a key given keepRaw although a handler writes it on an existing object, which needs strict
      • A new idempotent tool missing from idempotentToolCases in internal/server/handler/mutate_test.go; the idempotency tests skip it without failing
      • A new or renamed tool missing from the Tools section of README.md, the xmind-mcp-smoke-test skill, or the example map in cmd/gen-example (all three state the tool count; regenerate the map with make gen-example)
      • A tool whose handler returns jsonResult without handler.OutputSchema[...], missing from jsonOutputTools in internal/server/output_schema_test.go, or without a TestStructuredOutputMatchesSchema case for each success path
      • A whole-number parameter declared with mcp.WithNumber instead of mcp.WithInteger(..., mcp.Min(0)), or missing from nonNegativeIntParams in internal/server/tools_test.go
      • A new, renamed, or removed tool, or a workflow change, that leaves serverInstructions in internal/server/instructions.go stale (it may name only registered tools and must stay within 150 words)
      • An import between internal/xmind and internal/server/handler (see Layer Dependencies in AGENTS.md)
  • Invariants: If the project's agent instructions declare invariants (things that must never happen), any step that would break one is a blocker. These are always blockers:
    • Editing a migration that has already been applied
    • Weakening a test, constraint, or lint rule to make something pass
    • Removing, stubbing, or disabling existing behavior to clear an error without saying so explicitly
    • In this project, breaking a rule AGENTS.md declares in Fields to never modify, Topic IDs, the temp-then-swap write (Read/Write Lifecycle), or the no-write rule for idempotent tools (Argument extraction)

Clarity

  • Different-agent test: Could someone without the surrounding conversation context follow each step unambiguously?
  • Vague language: Flag uses of "appropriate," "relevant," "as needed," "properly" that lack specifics about what they mean concretely.
  • Actionability: Each step should be concrete enough to implement. Reserve "consider X" for places where real alternatives exist, and say what decides between them.

Completeness

  • Does the plan include a verification step (build/test/lint)?
  • Does it include documentation or config updates if structural changes are made?
  • Does it follow established patterns visible in the codebase?
  • Unknowns and assumptions: Does the plan rely on anything it has not confirmed? Flag assumptions that need checking, and missing context that prevents a confident review.

Severity levels

Classify each finding:

  • Blocker: Plan will fail or produce incorrect results if executed as-is. (e.g. wrong file path, missing step that breaks the build)
  • Warning: Plan will work but has a gap that could cause problems. (e.g. missing test step, vague wording an agent might misinterpret)
  • Suggestion: Plan is correct but could be improved. (e.g. reordering for efficiency, noting an edge case)

Task item review

If the plan has associated task items (YAML frontmatter todos, a checklist, or structured tasks), review each item for:

  • Incomplete: Lacks enough detail to implement unambiguously. A capable agent should not need to make assumptions.
  • Composite: Covers multiple distinct changes that could succeed or fail independently. Should be split.
  • Ambiguous: No clear completion criterion. You cannot tell when it is done.
  • Trivial: Will be completed automatically as a side-effect of another item, or is so self-evident it adds no value.
  • Misordered: Sequenced in a way that creates a blocker (a step depends on a later step).
  • Coverage gap: A significant section of the plan body has no matching task item.

Task item examples

Incomplete: "Update the parser" -- which parser? What change? Better: "Reject empty name values in parseConfig"

Composite: "Add the CSV exporter and write its tests" -- split into "Implement CsvExporter" and "Add tests for CsvExporter"

Ambiguous: "Clean up the code" -- no completion criterion. Better: "Extract duplicate path-validation into a shared helper"

Trivial: "Import the uuid package" -- this happens automatically when writing the code that uses it. Remove.

Misordered: "Add tests for CsvExporter" before "Implement CsvExporter" -- swap.

Output

  • Summarize what you verified and list findings with severity levels.
  • For errors, vagueness, or ambiguity: state the issue clearly, suggest a concrete fix. Do not edit the plan yet.
  • If required context is missing, ask targeted clarifying questions before proposing fixes.
  • If the plan is solid, say so briefly. No need to repeat it.

Applying changes (only after user approval)

Edit the plan document in-place. If the plan needs a fundamentally different approach (not just fixes), propose a replacement rather than a series of edits that would be harder to review. Proceed only after the user explicitly approves.

Related skills

  • After plan approval and implementation, run /verify-changes to confirm the build passes.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.