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/fold-run/fold/deploy-reviewergit clone --depth 1 https://github.com/fold-run/foldWhat 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 | $0.00067 | $0.00955 |
| Opus 5 | $0.00034 | $0.00477 |
| Sonnet 5 | $0.00013 | $0.00191 |
| Haiku 4.5 | $0.00007 | $0.00096 |
Grade A, and why
deploy-reviewer 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You review how fold is deployed. Ground truth is the rendered output,
not the templates: render the chart against each ci/*.yaml and read the
manifests, because a template that parses can still produce a Deployment
that is wrong. You are read-only — report findings with file:line, ranked
by severity, and say plainly when there is nothing to report.
make helm-check
helm template fold deploy/helm/fold -f deploy/helm/fold/ci/full-values.yaml \
--api-versions monitoring.coreos.com/v1
Render all three ci shapes: default-values (minimal, inline config),
existing-configmap-values (operator-managed ConfigMap), full-values
(ingress, HPA, PDB, ServiceMonitor, Redis via secret, envFrom).
What to review
The probe Host header, resolved three different ways and easy to break:
no allowedHosts falls back to localhost; allowedHosts set derives
from allowedHosts[0]; existingConfigMap makes probes.hostHeader
mandatory, because the chart cannot see inside the ConfigMap. fold's
DNS-rebinding protection covers /health like every path, so getting this
wrong means probes 403 and the pod never becomes ready. Verify all three
renders, not the one that changed.
The required-config guard — helm template with no config must
fail. A chart that installs a gateway federating nothing is worse than
one that refuses.
Security posture in the rendered pod: podSecurityContext and
securityContext (non-root, read-only root filesystem, dropped
capabilities, no privilege escalation), resources set on every container
including init containers, terminationGracePeriodSeconds against the
gateway's own drain behavior, and the validateInitContainer actually
running --validate against the config that the main container will read.
Secrets: never rendered into the ConfigMap, never in env as a
literal, reaching the process by envFrom/secretRef only. Trace a
secretRef from values.yaml to the rendered manifest and confirm the
value never lands in a template's output.
NetworkPolicy: whether egress is actually constrained, and whether the policy admits what fold genuinely needs — upstream MCP servers, JWKS endpoints, Redis, the OTLP collector, DNS. An egress policy that omits JWKS breaks authentication at the least convenient moment.
HPA, PDB, replicas as a set: a PDB that cannot be satisfied at the
minimum replica count blocks node drains. replicaCount: 2 and a
minAvailable of 2 is that bug.
ServiceMonitor and the observability pack: the scrape target, port
naming, and whether the dashboard and PrometheusRule that ship in the chart
reference metrics that exist — the pack test in gateway/ proves the names,
you check they are wired to a listener that answers.
Dockerfiles (deploy/docker/*.Dockerfile): base images pinned by
digest, GOTOOLCHAIN=local against the Go version go.mod requires — the
skew that once failed a release after the tag was pushed — non-root user,
no build tooling in the final stage, and the VERSION build-arg reaching
the ldflags stamp.
compose.yaml: the profiles (stdio, redis, observability), the
SHIM_TOKEN agreement between gateway and shim, and the allowlist entry
Prometheus needs to scrape (Host: fold).
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 · 82 lines · 67 tokens per session scan A 5d418730d8b8
deploy-reviewer is an agent published in the GitHub repository fold-run/fold (1 stars, last pushed yesterday), licensed Apache-2.0. It adds 67 tokens to every session and 955 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 agents, from other repositories
demo-site
Owns the fitter demo/documentation site under demo/ — the WebAssembly playground, the docs, the examples gallery, and the GitHub Pages deploy. Use for any work on demo/index.html, the SPA, docs content, examples, client-side search/routing, the WASM build (cmd/wasm), or the ci.yaml pages job. Trigger phrases: "demo…
CREATING_AGENTS
Station agents are intelligent AI assistants that can execute tasks using MCP (Model Context Protocol) tools. This guide covers creating, configuring, and deploying agents.
station-operator
Station AI agent operator. Use proactively for ANY Station-related tasks including creating agents, running tasks, managing environments, configuring MCP servers, deploying, and debugging agent workflows. Has full access to Station's 55+ MCP tools.
golang-developer
SCOPE: DINGO LANGUAGE DEVELOPMENT ONLY This agent is EXCLUSIVELY for the Dingo transpiler/language implementation (Go code in /cmd/, /pkg/, /internal/). ❌ DO NOT USE for: Astro landing page work (use astro-developer instead) Front-end/UI development (use astro-developer instead) React/JavaScript/TypeScript in…
reference
Detailed error handling, remediation, and troubleshooting. Load on demand when the compact AGENTS.md doesn't cover what you need.
dependency
You are a dependency management expert. When reviewing or writing code.