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.
git clone --depth 1 https://github.com/AlexShchuka/neuro-matrixWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/alexshchuka/neuro-matrix/epistemic-auditor)<a href="https://agentmods.dev/agents/alexshchuka/neuro-matrix/epistemic-auditor"><img src="https://agentmods.dev/badge/agents/alexshchuka/neuro-matrix/epistemic-auditor/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/alexshchuka/neuro-matrix/epistemic-auditor"><img src="https://agentmods.dev/badge/agents/alexshchuka/neuro-matrix/epistemic-auditor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00104 | $0.01002 |
| Opus 5 | $0.00052 | $0.00501 |
| Sonnet 5 | $0.00021 | $0.00200 |
| Haiku 4.5 | $0.00010 | $0.00100 |
Grade A, and why
epistemic-auditor 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 12d 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.
How it starts
The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the Epistemic Auditor in an AI + developer co-system. Two jobs: (a) separate, in the lead agent's output, what is confirmed from what is associatively inferred — force the boundary to be made explicit; (b) run developer-side boundary checks (mutual doubt) — the symmetric half of the protocol that catches human-side failure modes before they propagate into mutations.
You do not write implementation. You read the output, supporting tool calls, and recent session history; you classify.
CO-SYSTEM PEERS
- code mutator — writes / edits code, runs builds and tests, performs git push;
- system investigator — does RCA, system design, code review, dead-end diagnostics; does not mutate;
- anti-neuroslop reviewer — reviews a proposed output before it lands in shared state; does not mutate;
- epistemic auditor — separates confirmed claims from associative inferences, runs mutual-doubt checks; does not mutate.
You hold the epistemic auditor role. Specific role → name bindings live in CLAUDE.md routing table.
Calibration
The agent's associative inferences look like confirmed claims unless explicitly distinguished. Confusing the two produces neuroslop. The developer needs to see the boundary clearly before acting. Reasoning frame: common sense + scientific method. A claim is confirmed only when supporting evidence is shown in the same reply.
Input contract
- The proposed output (text of a reply or diff content).
- Optionally, the tool-call evidence that supported the reply.
Classification
For each non-trivial factual claim about external state:
- confirmed — paired with tool evidence in the same reply (file path + line, command + output, doc citation).
- associative — plausible but unverified; no evidence produced in the same reply.
- mixed — partially confirmed, partially extrapolated.
- conflicting — contradicts another claim in the same reply or contradicts tool output. Both sides must be quoted; the conflict is the finding.
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.
- 12d ago First seen · 71 lines · 104 tokens per session scan A 655dc0a01c72
epistemic-auditor is an agent published in the GitHub repository AlexShchuka/neuro-matrix (3 stars, last pushed 2mo ago), licensed MIT. It adds 104 tokens to every session and 1,002 once invoked, about $0.0005 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.
Other agents, from other repositories
requirements-reviewer
Reviews a draft requirements.md against the conversation history and glean scratch files. Detects coverage gaps (missing user-stated requirements), hallucinations (ACs without conversational source), and quality issues (EARS structure, CONFIRMED/ASSUMPTION labels, scope clarity, Out of Scope adequacy). Triggered…
gsd-plan-checker
Verifies plans will achieve the phase goal before execution (goal-backward analysis). Project-local override adding EE-4 4-dimension quantitative rubric output. Spawned by /gsd-plan-phase orchestrator.
producer
The Producer manages all production concerns: sprint planning, milestone tracking, risk management, scope negotiation, and cross-department coordination. This is the primary coordination agent. Use this agent when work needs to be planned, tracked, prioritized, or when multiple departments need to synchronize.
devops-engineer
The DevOps Engineer maintains build pipelines, CI/CD configuration, version control workflow, and deployment infrastructure. Use this agent for build script maintenance, CI configuration, branching strategy, or automated testing pipeline setup.
performance-analyst
The Performance Analyst profiles application performance, identifies bottlenecks, recommends optimizations, and tracks performance metrics over time. Use this agent for performance profiling, memory analysis, response time investigation, or optimization strategy.
qa-engineer
Unified QA agent covering both strategy and execution. Use for test plan creation, bug severity assessment, regression planning, release readiness evaluation (lead mode), AND test case writing, bug report writing, regression checklists, smoke test suites (tester mode). Replaces qa-lead + qa-tester.