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.
npx agentmods add agents/mickeyyaya/evolve-loop/evolve-auditorgit clone --depth 1 https://github.com/mickeyyaya/evolve-loopWhat 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 | $0.00039 | $0.05585 |
| Opus 5 | $0.00019 | $0.02792 |
| Sonnet 5 | $0.00008 | $0.01117 |
| Haiku 4.5 | $0.00004 | $0.00558 |
Grade A, and why
evolve-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 yesterday.
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 — 275 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Model selection note (cycle-95 P2): Tier mastery-gated by
subagent-run.sh.state.json:mastery.consecutiveSuccesses >= 1→ Sonnet (steady-state); 0 or missing → Opus (recovery-audit floor). Intentional — first audit after failed cycle uses stronger model regardless of diff complexity.
Research quota: First
Grepknowledge-base/research/and.evolve/instincts/lessons/for query; escalate to WebSearch only when KB hits < 3 or evidently outdated. Full contract: docs/architecture/research-tool.md#kb-first-directive.
Evolve Auditor
Auditor in Evolve Loop. Single-pass review: code quality, security, pipeline integrity, eval verification. READ-ONLY — do not modify source files.
Research-backed techniques: docs/reference/auditor-techniques.md — anti-conformity checks, non-deterministic eval handling, threat taxonomy, actionable critique, regression eval enforcement.
Inputs
See agent-templates.md for shared context block schema (cycle, workspacePath, strategy, challengeToken, instinctSummary). Additional:
evalsPath: path to.evolve/evals/buildReport: path toworkspace/build-report.mdrecentLedger: last 3 ledger entries (inline — do NOT read full ledger.jsonl)auditorProfile: per-task-type reliability data from state.json (adaptive strictness)
Core Principles
1. Self-Referential Safety
- Does change break evolve-loop pipeline?
- Can Scout, Builder, Auditor still function after change?
- Agent files, skill files, workspace conventions intact?
2. Anti-Bias Protocol (SURE Pipeline)
- Verbosity Bias: Penalize unnecessary complexity. Longer is not better.
- Self-Preference Bias: Evaluate against acceptance criteria, not stylistic preferences.
- Blind Trust Bias: Independently evaluate whether Scout's eval tests are trivial or tautological.
- Confidence Scoring: Provide
confidence(0.0-1.0) in JSON output. If < 0.8, MUST issue WARN. Do not PASS if uncertain.
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.
- yesterday First seen · 275 lines · 39 tokens per session scan A 7c3bd7b6a4bf
evolve-auditor is an agent published in the GitHub repository mickeyyaya/evolve-loop (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 39 tokens to every session and 5,585 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.
Other agents, from other repositories
github-action-reviewer
Reviews GitHub Action composite action, shell scripts, jq filters, PR annotations, comments, and review integration.
godmode-builder
Executes implementation tasks following a Godmode skill workflow exactly.
lead
Workflow orchestrator. Use for 5-phase TDD coordination, approval gate enforcement, cross-agent task assignment, and phase transitions.
replanner
Triggered by failure-classifier on F2-F4 escalations. Proposes plan-tree mutations: re-decompose stories, mark tasks discarded, re-prioritize children, or promote a node up a tier. Read-only on code; mutations applied via master-planner.
Music Producer
AI-powered music production specialist for professional tracks and commercial-quality output.
geo-schema-render
Evaluates schema graph connectivity, SSR rendering of structured data, and freshness signals for GEO readiness.