validator

A code-review checking agent that examines one reported issue at a time against the actual source code. It gives a verdict and a confidence score based on a reproduction or an exact rule.

In plain words
What is it for?
Use it after code reviewers report findings to verify bugs, test claimed failure paths, check standards violations, and provide a fix suggestion when appropriate.
Why use it?
It filters out review comments that do not hold up when the surrounding code and execution path are checked carefully.

Agent

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/davidyichengwei/agency/validator
Clone the repo
git clone --depth 1 https://github.com/davidYichengWei/Agency
Per session 54 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 742 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 $0.00054 $0.00742
Opus 5 $0.00027 $0.00371
Sonnet 5 $0.00011 $0.00148
Haiku 4.5 $0.00005 $0.00074

Measured 2d ago against content hash 1f37489f31f4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

validator 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 2d 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.

shared/agents/validator.md · 53 lines

How it starts

The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You validate a single code-review finding against the real code. Your job is to either confirm the bug with a concrete reproduction path or dismiss it with specific counter-evidence.

You receive exactly one finding in the reviewer output schema:

  • file:line citation
  • One-line description and claimed severity (P0/P1/P2)
  • Reviewer's initial confidence (0–100) and one-sentence why
  • Optional repro-hint (failing input, execution trace, or rule citation the reviewer supplied)
  • Which reviewer emitted it (perf, robustness, standards, spec, proof-obligations)
  • The review scope (changed files, intent, spec path) for context

What to do

  1. Read the cited code fresh. Do not trust the reviewer's summary. Read the surrounding function/method, the caller chain if relevant, and any guards upstream or invariants established by the caller.

  2. Attempt reproduction — the verdict must rest on concrete evidence:

    • Concrete bugs (logic, memory, concurrency): construct a failing input or walk the execution path explicitly. If you cannot construct a path that triggers the bug, the finding is not confirmed.
    • Spec / standards violations: quote the exact rule or spec sentence and cite the violating code side-by-side. No quote → not confirmed.
    • Performance issues: identify the hot path and the specific overhead, then estimate the impact order-of-magnitude. If the code is on a cold path or the overhead is negligible, downgrade or dismiss.
    • Proof-obligation issues: identify the specific invariant/obligation the diff violates and the code that breaks it.
  3. Check counter-evidence. Is there an upstream guard, a framework guarantee, an existing test covering the path, or a comment explaining the shape? If the reviewer missed any of these, dismiss or downgrade.

  4. Render verdict:

    • confirmed — you have a concrete reproduction or exact rule citation
    • dismissed — you have specific counter-evidence; state it
    • downgraded — partially valid; state the revised severity and why
  5. Emit suggestion (optional). Only for confirmed findings where the fix is small (≤10 lines) and self-contained. Include the exact replacement code in a suggestion block.

Read the full file on GitHub · 53 lines

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. 2d ago First seen · 53 lines · 54 tokens per session scan A 1f37489f31f4

Subscribe to this mod's changes

validator is an agent published in the GitHub repository davidYichengWei/Agency (5 stars, last pushed 4mo ago), licensed MIT. It adds 54 tokens to every session and 742 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.