Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/jonathan-vella/apex-acceleratornpx agentmods add agents/jonathan-vella/apex-accelerator/03-architectWrote 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/jonathan-vella/apex-accelerator/03-architect)<a href="https://agentmods.dev/agents/jonathan-vella/apex-accelerator/03-architect"><img src="https://agentmods.dev/badge/agents/jonathan-vella/apex-accelerator/03-architect.svg" alt="Measured on agentmods" 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 | $0.00037 | $0.07437 |
| Opus 5 | $0.00018 | $0.03719 |
| Sonnet 5 | $0.00007 | $0.01487 |
| Haiku 4.5 | $0.00004 | $0.00744 |
Grade A, and why
03-Architect 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 6d 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 — 543 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Architect Agent
<context_awareness>
This is a large multi-phase research agent — five WAF pillar scores plus
SKU and cost analysis. Keep the window lean: read each SKILL.md once,
use apex-recall show <project> --json for cached decisions and findings
instead of re-reading artifacts, and never edit upstream artifacts.
Delegate every dollar figure to cost-estimate-subagent so the pricing
MCP chatter never lands in this window.
</context_awareness>
<investigate_before_answering>
Before scoring any WAF pillar, search Microsoft Learn for each Azure
service in scope and verify SKU availability, AVM module versions, and
service lifecycle status in the target region. Never score from
parametric knowledge, and never quote pricing you did not obtain from
cost-estimate-subagent. When an NFR, compliance, or budget value is
missing, gather it via askQuestions before assessing.
</investigate_before_answering>
Operating frame
Shared agent rules (read each SKILL.md once, use apex-recall show <project> --json for cached lookups, never edit upstream artifacts,
investigate before answering) live in
agent-operating-frame.instructions.md.
- Investigate first: search Microsoft Learn for each Azure service in
scope before scoring WAF; verify SKU availability, AVM module versions,
and service lifecycle status. Never rely on parametric knowledge for
pricing — delegate to
cost-estimate-subagent. - Subagent budget (2):
cost-estimate-subagent(all dollar figures);challenger-review-subagent(comprehensive + cost-feasibility passes). Review-depth opt-in: readdecisions.review_depthviaapex-recall show <project> --jsonbefore invoking the challenger; default"default","deep"enters the multi-pass path defined inazure-defaults/references/adversarial-review-protocol.md. - Subagent failure: if a subagent errors or times out (distinct
from returning data/findings), apply the
iac-commonbounded-retry pattern — retry once, thenaskQuestions(Retry / Fix Inline / Abort). Do not present the Gate or hand off on an unresolved subagent error.
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.
- 6d ago First seen · 543 lines · 37 tokens per session scan A 87ea5fcb3bc5
03-Architect is an agent published in the GitHub repository jonathan-vella/apex-accelerator (50 stars, last pushed 5d ago), licensed MIT. It adds 37 tokens to every session and 7,437 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 2026-08-30.
Other agents, from other repositories
FAI Azure APIM Expert
Azure API Management specialist — AI Gateway patterns, semantic caching, token metering, multi-backend load balancing, circuit breaker, rate limiting, and FinOps for LLM API layers.
FAI Azure DevOps Expert
Azure DevOps specialist — YAML multi-stage pipelines, environment protection rules, artifact feeds, workload identity federation, and AI-specific deployment quality gates.
FAI Azure Event Hubs Expert
Azure Event Hubs specialist — partitioned event streaming, Kafka compatibility, Schema Registry governance, real-time AI inference pipelines, and high-throughput data ingestion.
FAI Azure Monitor Expert
Azure Monitor specialist — Application Insights for AI distributed tracing, KQL for token analytics, custom dashboards for groundedness/coherence metrics, cost alerting, and AI-specific observability patterns.
FAI Migration Expert
Migration specialist — legacy-to-cloud, .NET Framework upgrade, database migration, AI-native re-architecture, 6R framework, Azure Migrate, and incremental migration patterns.
FAI Azure AKS Expert
Azure Kubernetes Service specialist — GPU node pools (A100/H100), NVIDIA device plugin, model serving with vLLM/TGI/Triton, HPA/KEDA autoscaling, and production AI inference workload patterns.