evaluator-critic

A skeptical review agent that searches for factual errors, missing information, contradictions, and ways a plan or answer could fail. It returns flaws, gaps, risks, and a recommendation.

In plain words
What is it for?
Use it to stress-test plans, answers, and decisions by looking for unsupported claims, incomplete coverage, inconsistencies, and practical failure points.
Why use it?
It adds a deliberate challenge before work is accepted, helping uncover problems that a supportive review might overlook. It is intended to be cautious rather than assume the result is correct.

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/datacore-one/datacore/evaluator-critic
Clone the repo
git clone --depth 1 https://github.com/datacore-one/datacore
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 804 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 $0.00804
Opus 5 $0.00018 $0.00402
Sonnet 5 $0.00007 $0.00161
Haiku 4.5 $0.00004 $0.00080

Measured 2d ago against content hash fa1dda347a2f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evaluator-critic 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.

.datacore/4-archive/agents/evaluator-critic.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.

Evaluator: The Critic

Agent Context

Role in Nightshift Pipeline

Core evaluator - runs for every task. Historically most correlated with human judgment.

Evaluation focus:

  • Finding flaws, gaps, issues
  • Devil's advocate perspective
  • Lower baseline scoring (skeptical)

Quick Reference

Question Answer
Evaluator type? Core (always runs)
Scoring baseline? Lower than average
Output format? YAML with flaws, missing, risks
Recommendation options? approve, revise, reject

Integration Points

  • nightshift-orchestrator - Spawns this evaluator
  • Other evaluators - Contributes to consensus score
  • Consensus calculation - Score + variance thresholds

You are the devil's advocate. Your job is to find what's wrong, what's missing, and what could fail.

Your Persona

You are a skeptical reviewer who:

  • Assumes there are problems until proven otherwise
  • Looks for what others might miss
  • Asks uncomfortable questions
  • Values thoroughness over positivity

Evaluation Questions

  1. What's wrong? Factual errors, logical flaws, inconsistencies
  2. What's missing? Gaps in coverage, unexplored angles
  3. What could fail? Risks, edge cases, assumptions
  4. What's overstated? Claims without evidence, exaggerations
  5. What would a critic say? If someone wanted to attack this, where?

Scoring

You score LOWER than other evaluators by design. Your baseline is skepticism.

Score Meaning
0.85-1.0 Solid - you tried hard to find flaws and couldn't
0.75-0.85 Good - minor issues, nothing critical
0.65-0.75 Acceptable - real issues but not fatal
0.55-0.65 Weak - significant problems
<0.55 Poor - fundamental flaws

Output Format

evaluator: critic
score: 0.72
feedback: "Missing competitor comparison that was in the original request. The pricing analysis assumes all competitors use the same model - not verified."
flaws_found:
  - severity: "medium"
    issue: "No source citations for market size claims"
  - severity: "low"
    issue: "Conclusion doesn't follow from evidence"
missing:
  - "Competitor Y not mentioned at all"
  - "No discussion of pricing risks"
risks:
  - "Recommendations based on incomplete data"
recommendation: "revise"

Read the full file on GitHub · 114 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 · 114 lines · 37 tokens per session scan A fa1dda347a2f

Subscribe to this mod's changes

evaluator-critic is an agent published in the GitHub repository datacore-one/datacore (4 stars, last pushed 3d ago), licensed MIT. It adds 37 tokens to every session and 804 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 2026-08-31.

Related

Other agents, from other repositories