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 skills add AURORA-NEURO/aurora-agent --skill honest-labellinggit 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/honest-labelling)<a href="https://agentmods.dev/skills/aurora-neuro/aurora-agent/honest-labelling"><img src="https://agentmods.dev/badge/skills/aurora-neuro/aurora-agent/honest-labelling/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/honest-labelling"><img src="https://agentmods.dev/badge/skills/aurora-neuro/aurora-agent/honest-labelling.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 | $0.00083 | $0.00779 |
| Opus 5 | $0.00042 | $0.00390 |
| Sonnet 5 | $0.00017 | $0.00156 |
| Haiku 4.5 | $0.00008 | $0.00078 |
Grade A, and why
honest-labelling 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 — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Honest labelling
A system that tells you what it included is a retrieval library. A system that tells you what it left out, and whether that could have mattered, is worth trusting. These rules came out of the AURORA Agent workspace, where honest labelling is the product; they apply anywhere claims are produced.
The non-negotiables
- Zero influence is not unknown influence. "Provably cannot matter" and "nobody checked" are different states and must never share a representation. A single unknown-influence group voids a sufficiency claim.
- Unmeasured is not zero. A capability with no evidence is
Unmeasured, categorically distinct from measured-and-poor. There is noscore_or_zero. - A right answer from an incomplete basis is not a pass. Verify the inputs a decision was entitled to see before crediting the decision. A strategy must not be credited for guessing correctly from evidence it never had.
- Instance count is not benchmark count. Report independent equivalence classes; a million paraphrases are a robustness check, not a million benchmarks.
- Name what is not implemented. A missing capability that is stated is a limitation; one that is implied to exist is a lie. Keep an explicit list per module.
Make the rule unbreakable, not just tested
Where a rule can be made unrepresentable in the type system, make it unrepresentable. Patterns that shipped:
| Rule | Enforcement |
|---|---|
| A budget cannot be duplicated | the budget type does not implement Clone/Copy |
| An unmeasured capability has no score | private fields + one gated constructor |
| Provenance cannot be forged | fields private with an internal seal; serialize-only |
| Replay cannot fall through to live | the replay host has no live-source field, so no such branch exists |
| A state needs human approval | approve() is the only path to the approved type |
| Progression needs confirmation | the variant carries a token only the confirmation gate can mint |
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 · 67 lines · 83 tokens per session scan A 0ad64b632077
honest-labelling is a skill published in the GitHub repository AURORA-NEURO/aurora-agent (1 stars, last pushed yesterday), licensed Apache-2.0. It adds 83 tokens to every session and 779 once invoked, about $0.0004 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
halopsa
Every HaloPSA, HaloITSM and HaloCRM feature, plus a local SQLite store and cross-entity views the API can't return. Trigger phrases: triage my Halo queue, check SLA breaches in HaloPSA, who is overloaded in Halo, client card for Acme in Halo, Halo contract burn-down, what changed in Halo since this morning, find time…
veeam
Use when the user asks to check Veeam backup health across customers, find failing or stale backup jobs, surface workloads past their RPO, triage VSPC alarms, or report per-tenant license usage from a Veeam Service Provider Console. Syncs every tenant into a local SQLite mirror so it answers multi-tenant backup…
cipp
First single-binary CLI for CIPP - offline SQLite store, fleet posture analytics, and cross-tenant fan-out no other CIPP tool has. Trigger phrases: check MFA across all my tenants, find unused M365 licenses, offboard users from a CSV in CIPP, which tenants drifted off baseline, list CIPP tenants, use cipp, run…
appdirect
Every documented AppDirect marketplace operation in one binary, plus offline sync and billing-reconciliation joins. Trigger phrases: reconcile appdirect billing, which appdirect payments failed this week, what changed in appdirect subscriptions, show my appdirect pipeline, appdirect company 360, use appdirect, run…
connectwise-manage
Every ConnectWise PSA workflow from the terminal - with a typed conditions query builder, offline SQLite sync, and cross-entity views (unbilled work, account 360, board triage) the PSA web UI can't give you. Trigger phrases: list connectwise tickets, triage the help desk board, find unbilled time in connectwise…
immybot
Every ImmyBot endpoint typed, plus a local SQLite mirror that answers the cross-tenant questions the web UI cannot. Trigger phrases: what failed in last night's maintenance window, which tenants are still on an old version of chrome, why didn't this computer get the deployment, what changed in the fleet since…