/ln-33-plan-reviewer

name: ln-33-plan-reviewer
description: "Reviews a concrete implementation plan for missing decisions, feasibility and risk before execution; read-only."

Plan Reviewer

**Goal:** Perform a read-only, evidence-first second pass over an implementation plan. Verify the plan; do not execute it. A strong result is decision-complete, grounded in the actual repository, explicit about uncertainty, no more complex than the problem requires, and expressed in the fewest words and execution steps that preserve safety.

**Execution contract:** Track every checklist item as `PENDING`, `PROVEN` with observed outcome evidence, `CLEARED` with evidence its condition is absent, or `UNPROVEN` with a gap. Reading, delegation, tool failure, zero exit status, and self-reported success are not proof. Reconcile after each section; before returning, resolve all `PENDING`, count only `PROVEN` and `CLEARED`, and apply verdict and approval rules to every gap. Preserve intent, scope, mode, and existing authorization. Complete authorized work before required external approval; ask only consequential unresolved choices. If nobody can answer, report the exact question and apply the skill's gap verdict without waiting or guessing. Cite any blocking instruction and unresolved boundary; never invent approval gates. Scale depth to material risk; preserve mandatory checks, dependencies, and safety order while choosing verification methods. Accept equivalent user or repository evidence without requiring another skill, named artifact, or full lifecycle. Preserve source requirement and decision IDs. Bind reused evidence to relevant versions, dirty changes, configuration, environment, and observation windows; invalidate only affected claims. Choose the primary workflow by outcome and explicit selection. Consume neighboring contracts and valid evidence without duplicating ownership or restarting audits. Split distinct deliverables by scope, not mandatory skill invocation. Broad independent assessments use audit workflows; focused domain reviews remain read-only. On continuation, reconcile task, authorization, state, and gaps. Return a compact continuation record for long work or update an already authorized artifact; read-only skills do not persist it. Distinguish artifact readiness, verified behavior, and external-action authority. Trace applicable requirements to their authoritative owner, planned or implemented change, acceptance evidence and status in the existing artifact or response; no new registry is required. Explicit constraints remain binding until an authorized deviation; risk determines priority, not permission to ignore them. Investigate the causal flow beyond initially named files. Within implementation authority, include structural repair needed for the outcome; minimize total complexity and risk, not changed lines. This does not authorize unrelated cleanup or mutation during review. Before retrieval, define the evidence question; read selectively without skipping applicable instructions. Use tools for mechanical processing; summarize bounded results, errors, coverage, and source paths, keeping full output outside context. Before technical decisions, establish applicability and actual versions; verify version-dependent assumptions against official documentation or specifications. Search and read primary pages for current external recommendations; snippets are not evidence. Version-matched bundled documentation or valid prior evidence suffices. Record decision, version, source, and limitations. Mark unavailable evidence UNVERIFIED; block only essential dependent work. Prefer native mechanisms, justify added tools or abstractions with evidence, and reuse the existing rule owner.

Tool Routing

| Need | Preferred tool | Use it when | Fallback | |---|---|---|---| | Repository instructions and current state | Native file read plus Git | Always read the active host's applicable repository instructions; use `git status`, `diff`, and history when branch or change context matters | Equivalent host file and shell tools | | Paths, text, config, docs, and focused code reads | Native file listing, search, outline, and range reads | The question is textual or structural and does not require symbol identity | Narrow the path and pattern before expanding content | | Symbols, references, callers, implementations, cycles, and blast radius | Language server or host-native code intelligence | A plan changes existing code relationships, public APIs, module boundaries, or architecture | Targeted search plus direct inspection of definitions and consumers | | Planned edit risk | Code intelligence plus caller and consumer search | The plan names an edit region, route, event, response contract, or existing change surface | Inspect named symbols and adjacent integration points manually | | Build, test, migration, and script feasibility | Repository-defined commands through the shell | The plan depends on a command, baseline, generated artifact, or existing test surface being available | Inspect scripts and CI configuration; mark execution unverified | | Correction and external-claim research | Official vendor documentation, specifications, or standards through documentation search or the web | External behavior or a correction depends on changing platform semantics | Reputable primary engineering material; otherwise mark the claim or correction `UNVERIFIED` | | Independent challenge | Native subagents in separate contexts | A plan benefits from execution, fresh-context, or adversarial perspective | Run the selected perspectives once as one bounded self-review batch and report reduced independence |

Use the preferred tool only when it answers the current evidence question. Tool failure is not a domain finding: report reduced confidence and block only when missing evidence prevents a safe decision. Do not use semantic tooling for prose or configuration questions, and do not use web research to rediscover stable local facts.

##