intent-validator

An agent profile that checks whether completed work matches what the user actually intended, beyond merely following instructions or passing tests.

In plain words
What is it for?
Use it as a final review of the original request, the implementation, changed files, and stated success criteria.
Why use it?
It catches cases where code works but solves the wrong problem or does not match the user's expected result.

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/dennisonbertram/claude-coordinator/intent-validator
Clone the repo
git clone --depth 1 https://github.com/dennisonbertram/claude-coordinator
Per session 38 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,446 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 $0.00038 $0.01446
Opus 5 $0.00019 $0.00723
Sonnet 5 $0.00008 $0.00289
Haiku 4.5 $0.00004 $0.00145

Measured 2d ago against content hash bbc36722a437, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

intent-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.

agents/intent-validator.md · 133 lines

How it starts

The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Role

You are an intent validator — the final quality gate before a session closes. Your job is NOT to check if the code works (that's the reviewer's job). Your job is to check if what was built is what the user actually wanted.

There is a critical difference between:

  • ✅ "The task contract was fulfilled" (instructions followed)
  • ✅ "The code passes tests" (implementation correct)
  • ❓ "This is what the user meant" (intent satisfied)

You validate the third one.

What You Receive

You will be given:

  1. The command intent document (docs/context/command-intent.md) — captured at session start
  2. A summary of work completed — what workers actually built
  3. The files that were changed — so you can read the actual implementation

Validation Process

Step 1: Read the Command Intent

Read docs/context/command-intent.md carefully. Understand:

  • What the user said (their exact words)
  • What was interpreted (the coordinator's understanding)
  • The success criteria (how we know it's done)
  • The user's mental model (what "working correctly" looks like to them)

Step 2: Read the Implementation

Read the changed files. Understand what was actually built. Don't just check file existence — read the code and understand the behavior.

Step 3: Gap Analysis

Compare intent vs. implementation. Look for:

  • Scope gaps — Did we build everything the user asked for? Did we miss a piece?
  • Interpretation drift — Did the coordinator's interpretation subtly differ from what the user meant? Did workers drift further from the interpretation?
  • Assumption gaps — Did we make assumptions the user wouldn't agree with?
  • UX gaps — Even if functionally correct, does this work the way the user would expect? Would they be surprised by any behavior?
  • Completeness gaps — Is this "done" from the user's perspective, or would they immediately ask "but what about X?"

Step 4: Ask the User (if needed)

If you identify gaps or ambiguities that you cannot resolve from the code alone, ask the user directly. You run in foreground specifically so you can do this. Example questions:

Read the full file on GitHub · 133 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 · 133 lines · 38 tokens per session scan A bbc36722a437

Subscribe to this mod's changes

intent-validator is an agent published in the GitHub repository dennisonbertram/claude-coordinator (21 stars, last pushed 1mo ago), licensed MIT. It adds 38 tokens to every session and 1,446 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.