Borrowing it
Nothing to install: this file belongs to AURORA-NEURO/aurora-agent. 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/AURORA-NEURO/aurora-agent/main/.agents/skills/check-parity/SKILL.mdgit clone --depth 1 https://github.com/AURORA-NEURO/aurora-agentWrote 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/skills/aurora-neuro/aurora-agent/check-parity)<a href="https://agentmods.dev/skills/aurora-neuro/aurora-agent/check-parity"><img src="https://agentmods.dev/badge/skills/aurora-neuro/aurora-agent/check-parity/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/skills/aurora-neuro/aurora-agent/check-parity"><img src="https://agentmods.dev/badge/skills/aurora-neuro/aurora-agent/check-parity.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 | $0.00916 |
| Opus 5 | $0.00028 | $0.00458 |
| Sonnet 5 | $0.00011 | $0.00183 |
| Haiku 4.5 | $0.00006 | $0.00092 |
Grade A, and why
check-parity 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 10d 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.
Check cross-language parity
Certificates hash canonical bytes. If two implementations disagree, a certificate produced by one cannot be replayed by the other, and the guarantee the whole scheme rests on is gone.
Three implementations must agree on the reference certificate:
c0da17ffc80465258345c8a538171bfd868100cd883e9a20780a0dc5477e7ea4
Run it
python tools/regenerate_golden.py
git diff --exit-code fixtures/
cargo test -p bioprism-ids -p bioprism-fiber -p bioprism-store --offline
A non-empty git diff means the CPython reference itself moved. bioprism-store's suite is what
proves the eager and indexed backends agree. CI runs all of this as its own job.
Why it is fragile, so you know what to suspect
Matching CPython required reproducing three behaviours a natural Rust port gets wrong.
Float formatting. CPython switches to exponential notation at a different threshold than Rust
and zero-pads the exponent to two digits. python_repr_f64 in crates/ids/src/canonical.rs
reimplements CPython's rule; tools/gen_python_ground_truth.py regenerates the expectation table.
Float parsing. serde_json's default parser is not correctly rounded and disagrees with both
CPython and native Rust near the subnormal boundary and at 2^53 — 2.2250738585072011e-308 lands
one ULP high, 9007199254740993.0 rounds the wrong way. The workspace enables the
float_roundtrip feature to fix it. If that feature is dropped, parity breaks silently and the
golden fixtures will not catch it, because they contain no such values. The regression test
float_parsing_agrees_with_native_rust_and_cpython is the only guard.
Object iteration order. The reference builds leakage witnesses in document order, so serde_json
runs with preserve_order. Canonical output sorts keys regardless, so digests are unaffected — but
witness list order is not.
Non-finite floats, which are a hole rather than a fixed behaviour. CanonicalError::NonFiniteNumber
exists and cannot fire on the path every producer uses: serde_json::to_value maps NaN, +inf
and -inf to Value::Null before the encoder sees a number, so all three hash as a stated null.
They do not collide with an absent key — that distinction survives — but a corrupt number is
silently laundered into a stated null, which is the wrong direction to fail in here. CPython's
json.dumps meanwhile emits the bare token NaN, which is not JSON and which serde would refuse to
parse, so the two languages disagree about what a non-finite float is and agree today only because
no producer emits one. crates/ids/tests/non_finite_floats.rs pins all of it. If you add a float
field to a hashed format, validate it at the producer — bioprism_biolang::BioState::validate is
the worked example.
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.
- 10d ago First seen · 71 lines · 57 tokens per session scan A 62f8a1fe665b
check-parity is a skill published in the GitHub repository AURORA-NEURO/aurora-agent (1 stars, last pushed today), licensed Apache-2.0. It adds 57 tokens to every session and 916 once invoked, about $0.0003 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 skills, from other repositories
blazemeter-performance-testing
Comprehensive guide for BlazeMeter Performance Testing, including load configuration, reporting, JMeter configuration, Taurus, scenarios, and advanced features. Use when working with Performance tests for (1) Configuring load settings and distribution, (2) Creating and running tests (JMeter, Browser, URL/API…
blazemeter-api-reference
Comprehensive reference for BlazeMeter REST APIs, including authentication, identifiers, and API endpoints. Use when working with BlazeMeter APIs for (1) Understanding BlazeMeter REST API structure, (2) Authenticating API requests, (3) Obtaining identifiers (Workspace ID, Project ID, Test ID, etc.), (4) Using Test…
blazemeter-api-monitoring
Comprehensive guide for BlazeMeter API Monitoring, including test creation, configuration, scripting, integrations, notifications, and management. Use when working with API Monitoring tests for (1) Creating and configuring API tests, (2) Writing custom scripts (Initial, Pre-request, Post-response), (3) Integrating…
blazemeter-getting-started
Getting started guides for BlazeMeter, including onboarding, migration guides, continuous testing journey, glossary, and mobile testing. Use when getting started with BlazeMeter for (1) Navigating BlazeMeter University onboarding, (2) Migrating from Runscope or JMeter, (3) Understanding the continuous testing journey…
blazemeter-integrations
Comprehensive guide for BlazeMeter Integrations, including APM tools, CI/CD pipelines, and development tools. Use when working with integrations for (1) Integrating APM tools (AppDynamics, Datadog, New Relic, CloudWatch, DX APM, Dynatrace, Delphix), (2) Integrating CI/CD tools (Jenkins, GitHub Actions, GitLab CI/CD…
blazemeter-service-virtualization
Comprehensive guide for BlazeMeter Service Virtualization, including virtual services, transactions, templates, and management. Use when working with Service Virtualization for (1) Creating virtual services and transactions, (2) Managing services (clone, export/import, rename/delete), (3) Using templates and…