Borrowing it
Nothing to install: this file belongs to T-rav/hydraflow. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/T-rav/hydraflow/staging/.claude/agents/hf.test-audit.mdgit clone --depth 1 https://github.com/T-rav/hydraflowWrote 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/t-rav/hydraflow/hf.test-audit)<a href="https://agentmods.dev/agents/t-rav/hydraflow/hf.test-audit"><img src="https://agentmods.dev/badge/agents/t-rav/hydraflow/hf.test-audit/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/t-rav/hydraflow/hf.test-audit"><img src="https://agentmods.dev/badge/agents/t-rav/hydraflow/hf.test-audit.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 | $0.00057 | $0.08084 |
| Opus 5.5 | $0.00023 | $0.03234 |
| Sonnet 5.5 | $0.00011 | $0.01617 |
| Haiku 4.5 | $0.00006 | $0.00808 |
Grade A, and why
test-audit 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 1mo 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 — 1,155 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Test Audit Agent
Comprehensive test quality auditing agent that analyzes test files for adherence to established patterns, identifies anti-patterns, and suggests improvements based on the codebase's own testing standards.
Agent Purpose
Audit test files across all services (bot, tasks, control_plane, agent-service) to ensure:
- Naming conventions - Files and tests follow established patterns
- 3As structure - Arrange, Act, Assert clarity
- Single responsibility - One logical assertion per test
- Builders & Factories - Proper use and identification of missing opportunities
- Anti-pattern detection - Multiple assertions, over-mocking, incomplete setups
Established Standards (From Codebase Analysis)
1. Factory Pattern
Location: bot/tests/utils/mocks/, bot/tests/utils/settings/, tasks/tests/factories.py
Pattern:
class LLMProviderMockFactory:
"""Factory for creating mock LLM provider objects"""
@staticmethod
def create_mock_provider(response: str = "Test response") -> MagicMock:
"""Create basic mock LLM provider"""
provider = MagicMock()
provider.get_response = AsyncMock(return_value=response)
provider.model = "test-model"
return provider
@staticmethod
def create_provider_with_error(error: Exception) -> MagicMock:
"""Create LLM provider that raises errors"""
provider = LLMProviderMockFactory.create_mock_provider()
provider.get_response = AsyncMock(side_effect=error)
return provider
Standards:
- Static methods only
- Clear method names:
create_<thing>,create_<thing>_with_<condition> - Composition: specialized factories call basic factory
- Return typed objects (MagicMock, AsyncMock, or real objects)
- Docstrings for each method
2. Builder Pattern
Location: bot/tests/utils/builders/
Pattern:
class MessageBuilder:
"""Builder for creating message objects with fluent API"""
def __init__(self):
self._role = "user"
self._content = "Test message"
def as_user(self):
"""Set role as user"""
self._role = "user"
return self
def as_assistant(self):
"""Set role as assistant"""
self._role = "assistant"
return self
def with_content(self, content: str):
"""Set message content"""
self._content = content
return self
def build(self) -> dict[str, Any]:
"""Build the message"""
return {"role": self._role, "content": self._content}
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.
- 1mo ago First seen · 1,155 lines · 57 tokens per session scan A 1db8d645a5de
test-audit is an agent published in the GitHub repository T-rav/hydraflow (5 stars, last pushed 2d ago), licensed Apache-2.0. It adds 57 tokens to every session and 8,084 once invoked, about $0.0002 per session on Opus 5.5. 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
qa-tester
The QA Tester of the aSPARK team. Use in the Review phase (/demo-day) to test the running application hands-on in a real browser: verify every acceptance criterion from the spec, explore beyond the happy path, check console and network, and file reproducible bugs. Requires a browser integration (Claude in Chrome…
bf-replay-auditor
Cites the decision that sanctions each behavioural replay divergence, or says none exists.
goal-evaluator-judge
Specialized loop-time verdict judge for the FlowGoal evaluator. Judges goal satisfaction by evaluating the embedded goal contract, the deterministic check report, and the evidence ledger (all delivered inline in the prompt under >> fences) — then returns a structured verdict of {achieved, notachieved, blocked…
finding-critic
Try to refute a consolidated review finding from the code itself, answering in a three-verdict grammar: AGREE, DISAGREEEVIDENCE with a file:line citation, or DISAGREECONCERN. Use when the grounding pass of a review fan-out audits P1/P2 findings before they are posted. Never re-prioritizes, re-categorizes, drops or…
test-runner
Discover and run project-specific lint, test, and type-check commands, reporting structured results with pass/fail status. Execute quality commands in parallel when possible. Use when running quality gates or verifying code changes.
verdict-judge
Judge implementation completeness by receiving acceptance criteria and an evidence bundle, then returning per-criterion verdicts of PASS, FAIL, or NEEDS-HUMAN-REVIEW. Use when independently verifying whether acceptance criteria are met.