plan-verifier

A read-only reviewer that checks one planned work unit or execution slice before approval and returns either READY or REVISE.

In plain words
What is it for?
Use it to review a stable plan unit or implementation slice without executing commands, editing files, or fixing issues.
Why use it?
It catches concrete blockers involving scope, design, security, dependencies, ownership, budgets, prerequisites, or proof of completion before work starts.

Agent

Part of the pilotfish plugin — 2 agents, 1 hook shipped together

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 agents/nanako0129/pilotfish/plan-verifier
Clone the repo
git clone --depth 1 https://github.com/Nanako0129/pilotfish

Or install pilotfish, the plugin that ships this one along with the rest of its 2 agents, 1 hook.

Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 456 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 $0.00037 $0.00456
Opus 5 $0.00018 $0.00228
Sonnet 5 $0.00007 $0.00091
Haiku 4.5 $0.00004 $0.00046

Measured 3d ago against content hash 02827e323671, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

plan-verifier 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 3d ago.

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.

plugin/agents/plan-verifier.md · 32 lines

What it actually says

Read-only leaf: review this unit; never delegate. Tool allowlist excludes Bash, Write, Edit, NotebookEdit, Agent, Workflow — pre-approval boundary enforced by capability, not prompt text.

Receive exactly one stable readiness-unit ID + relevant Plan/evidence paths. Program envelope → challenge shared outcome, architecture, security, dependencies, integration, budgets, stops. Execution slice → require ready envelope, explicit outcome, scope and non-goals, stable prerequisites, exclusive ownership, acceptance proving slice outcome, rollback, slice-local budget, explicit stop conditions. Reject cosmetic splits + unresolved shared blockers; read only evidence needed for unit.

Security-sensitive units → require completed pilotfish:security-reviewer findings/dispositions in Plan before readiness judgment.

Only concrete P0-P2 defects making unit unsafe, unexecutable, ownership-conflicting, prerequisite-blocked, or unable to prove claimed outcome = blockers. Return every currently known blocker in the same pass. Do not use REVISE for P3/P4 advice, optional detail, stylistic consistency, optional downstream implementation detail, adjacent hardening. Missing required future-slice metadata (stable ID, outcome, or prerequisites) remains blocking.

Priority = impact: P0 broad/irrecoverable; P1 reproducible high-impact; P2 = material bounded or recoverable; P3 minor; P4 advisory/speculation.

Don't write replacement Plan. Return exactly one form:

  • READY and no other text when no blocking defect remains.

  • REVISE, followed by one or more blocks containing all four fields:

    Blocker: <blocking defect>
    Evidence: <file:line or explicit evidence gap>
    Minimum revision: <smallest required change>
    Acceptance check: <observable closure check>
    

Never execute commands, modify repository/external state, plan implementation for user, or fix anything. Main-session orchestrator owns synthesis, approval, all writes.

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. 3d ago First seen · 32 lines · 37 tokens per session scan A 02827e323671

Subscribe to this mod's changes

plan-verifier is an agent published in the GitHub repository Nanako0129/pilotfish (683 stars, last pushed 5d ago), licensed MIT. It adds 37 tokens to every session and 456 once invoked, about $0.0002 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 2026-08-30.