engine

A private review engine that examines changes made during a coding session and returns structured findings as data. It is designed to be called by other review commands or programs, not directly by users.

In plain words
What is it for?
Use it as an internal step in code-review pipelines to analyze a session's changes, run registered checks, verify findings, and return a review-results object.
Why use it?
It gives automated review workflows a consistent format for collecting and checking findings without changing files or interrupting the user.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/apurvbazari/claude-plugins/engine
Any agent
npx skills add ApurvBazari/claude-plugins --skill engine
Clone the repo
git clone --depth 1 https://github.com/ApurvBazari/claude-plugins

Made for: Claude Code, Codex.

Per session 61 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,074 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00061 $0.03074
Opus 5 $0.00030 $0.01537
Sonnet 5 $0.00012 $0.00615
Haiku 4.5 $0.00006 $0.00307

Measured yesterday against content hash 25a5b7243102, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

engine scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

lens/skills/engine/SKILL.md · 114 lines

How it starts

The opening of the file, as written. The whole thing — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Engine — the data-only review core

Produce a review-findings object (per ../../schemas/review-findings.schema.json) and RETURN it to the caller. Write no files; ask no questions. Read references/engine-api.md (the declared input/return surface) + references/pipeline.md + references/finder-registry.md.

Progress tracking (optional — only when handed taskIds)

If the caller passed taskIds in args (the standalone /lens:review path hands { scope, intent, analyze, verify }), mark each handed-in stage task in_progress when you enter that stage and completed when you leave it, via TaskUpdate keyed on the given taskId. If taskIds is absent (orchestrator/compute-only callers, vicario reuse, any non-standalone invocation), take no task action — behavior is unchanged and byte-identical to the data-only contract. Never create tasks. The finder and verifier subagents you dispatch are task-blind — they emit findings only and never touch the task list.

Step 0: Pre-flight

Pre-flight runs before Step 1, ahead of its empty-scope short-circuit: an empty diff carrying a malformed 1.5.0 input returns the error envelope, not emptyScope:true. Three jobs, then the stages below run.

1. Validate the 1.5.0 keys. Each is checked against the shape declared in references/engine-api.md § lens:engine — inputs, and a violation raises E_INVALID_INPUT:

  • modelPolicyits top-level type first (an object, after the defensive string parse), then its four-key shape, its key types, and the accepted role names. The order is load-bearing: the shape, type and role checks are all key-scoped, so a policy that is not an object has no keys for any of them to judge and would pass all three — and be discarded silently, taking the deny gate with it.
  • verifyVotes — an integer inside the range 1..5.
  • model/effort on every call-time injectedFinders record — the declared vocabularies. A file-registered .claude/lens/settings.md record never raises here; its malformed model/effort is normalized-or-dropped here in pre-flight (references/pipeline.md §3) rather than at any stage below, because the plan resolved next is frozen from what survives — and because a human's project config is not the caller's to fix mid-call.

Read the full file on GitHub · 114 lines

Files

What ships with it

5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 114 lines · 61 tokens per session scan A 25a5b7243102

Subscribe to this mod's changes

engine is a skill published in the GitHub repository ApurvBazari/claude-plugins (0 stars, last pushed 25d ago), licensed MIT. It adds 61 tokens to every session and 3,074 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.