cuora
/cuora
Materialize secrets via the cuora CLI — cuora:// refs, env-var injection with `run`, stdin-only writes, --json surfaces, advisory guardrails. Trigger when a task needs API keys, tokens, or credentials in code, config, or child processes.
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 2
No signals detected. Repository changed on 13.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Materialize secrets via the cuora CLI — cuora:// refs, env-var injection with `run`, stdin-only writes, --json surfaces, advisory guardrails. Trigger when a task needs API keys, tokens, or credentials in code, config, or child processes.
Original author description
Materialize secrets via the cuora CLI — cuora:// refs, env-var injection with `run`, stdin-only writes, --json surfaces, advisory guardrails. Trigger when a task needs API keys, tokens, or credentials in code, config, or child processes.
How it works
- The one rule
- Bootstrap
- Discovery — prefer asking the CLI over reading docs
- The ref model
- Verb cheat sheet — by intent
- Writing secrets — stdin only
- Config shape (cuora.toml, searched ./ → ~/.config/cuora/)
- Cloud vaults (kind = "cf")
- Audit
SKILL.md preview
---
name: cuora
description: Materialize secrets via the cuora CLI — cuora:// refs, env-var injection with `run`, stdin-only writes, --json surfaces, advisory guardrails. Trigger when a task needs API keys, tokens, or credentials in code, config, or child processes.
metadata:
version: "1.0.0"
---
# cuora — secrets broker for agents
`cuora` removes the plaintext handoff. Code and config carry
`cuora://` references; the CLI resolves them and injects values as
env vars at run time. **You never see the value — that's the
point.**
## The one rule
**Never read a secret into context.** Don't `cuora get`, don't
`cuora env`, don't cat vault files, don't ask the human to paste
the value. Consume secrets by running the thing that needs them:
```bash
cuora run -- 'curl -H "Authorization: Bearer $STRIPE_KEY" https://api.stripe.com/v1/charges'
```
`run` injects declared secrets into the child's environment and
exits with the child's code. The value lives only in that
process — not in your output, not in logs, not in history.
## Bootstrap
```bash
command -v cuora # missing? it's a tamias pack — from a cuora checkout:
cargo xtask pack && tamias install --allow-drivers ./libraries/cuora
cuora init # generates ~/.config/cuora/identity.txt (0600)
```
## Discovery — prefer asking the CLI over reading docs
```bash
cuora manifest # verbs, json shapes, env vars, audit contract (JSON)
cuora schema # JSON Schema for cuora.toml
cuora <verb> --help # every verb, self-documenting
```
`--json` is honored on every verb; errors come back as
`{"error": "…"}` on stderr with a non-zero exit.
## The ref model
- `cuora://ns/name` names a logical secret — env is **context**, not
part of the name: `--env prod` / `CUORA_ENV` / `[envs]` config
pick `dev`/`staging`/`prod` (default `dev`).
- `cuora://prod/ns/name` **pins** the env regardless of context.
- `[secrets] VAR = "cuora://ref"` in cuora.toml declares what `run`
may inject — least privilege, `cuora check`-able.
## Verb cheat sheet — by intent
| Intent | Command |
| --- | --- |
| Store a secret | `printf '%s' "$val" \| cuora put cuora://ns/name` |
| Read a value (escape hatch) | `cuora get cuora://ns/name` — **avoid**; see the one rule |
| List refs (names only) | `cuora list [--env prod] [--json]` |
| Remove / rotate | `cuora rm <ref>` · `printf … \| cuora rotate <ref>` |
| Run with secrets (the way to consume) | `cuora run -- prog arg…` (argv exec) · `cuora run -- 'shell line'` (bash -c) |
| Ad-hoc injection | `cuora run -s VAR=cuora://ref -- cmd` (repeatable) |
| Exports for a human shell | `cuora env [-s VAR=ref]` → `export VAR='…'` lines for `eval` |
| Push to a write-only sink | `cuora deploy <ref> [--target NAME]` |
| Validate config + surface | `cuora check [--json]` — exit 1 on findings |
| Bootstrap | `cuora init` |
Two `run` exec modes: **one arg** after `--` → `bash -c` (pipes,
quoting); **several args** → `argv[0]` exec'd directly, no shell —
prefer argv when the command has arguments.
## Writing secrets — stdin only
Values **never** go on the command line (args land in shell
history and `/proc`). Pipe them:
```bash
printf '%s' "$STRIPE_KEY" | cuora put cuora://billing/stripe
pass show stripe/test | cuora put cuora://billing/stripe --env dev
```
`put`/`rotate` strip one trailing newline; the rest is verbatim.
Interactive humans get a hidden prompt instead.
## Config shape (`cuora.toml`, searched ./ → ~/.config/cuora/)
```toml
[vaults.local] # the built-in age vault; entry only overrides
kind = "age"
[vaults.cloud] # a remote vault — self-hosted CF Worker + D1
kind = "cf" # holding per-secret age ciphertext
url = "https://…workers.dev"
token_env = "CUORA_CF_TOKEN" # env var NAME — the token is never in toml
[envs] # env → vault routing; unlisted envs → "local"
prod = { vault = "cloud" }
[secrets] # the declared surface run/env may inject
STRIPE_KEY = "cuora://billing/stripe"
[deploy.prod] # write-only sinks `cuora deploy` pushes to
kind = "cf-secrets-store" # (value goes in, never comes back)
account_id = "…"
store_id = "…"
[guardrails] # advisory: accident prevention, NOT policy
deny = ["get"] # deny wins; default = "allow" for unlisted
```
If a guardrail denies your op, that's the owner saying "agents
don't do this" — don't work around it; use the allowed verb or
surface the denial to the human.
## Cloud vaults (`kind = "cf"`)
A `cf` vault talks HTTPS to a self-hosted Cloudflare Worker + D1
(deployed from `crates/cuora-vault-cf/worker/`). Same verbs, same
identity — values are age-encrypted client-side and the store holds
ciphertext only. Two requirements beyond `local`:
- the worker must be **deployed** — an unreachable `url` makes
every verb fail with an I/O error (`cuora check` reports the vault
`ok: false`, never a panic);
- the bearer token must be in the env var `token_env` names
(`CUORA_CF_TOKEN` by default) — ask the human for it, don't
invent one.
`cuora deploy cuora://ref [--target NAME]` pushes a *stored* value to
a write-only sink (`[deploy.NAME]`, e.g. a CF Secrets Store). The
value crosses to the sink in plaintext and can never be fetched
back — the vault stays the source of truth. It's a push, not a
sync: use it when a runtime must consume the secret directly from
Cloudflare's own store.
## Audit
Every materialization and every denial/failure emits one ndjson
event on **stderr**: `{ts, actor, op, rref?, env,
outcome: ok|denied|error}`. `CUORA_ACTOR`/`--actor` sets
attribution. Don't parse values out of stderr — there are none;
the wire carries metadata only.
Versions
The database has no other skills named cuora.
Database snapshot: 13.11.2026 · package version 7b655c771112 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100onepassword-secrets
Injects and audits 1Password secrets via the op CLI using op:// references, prefers op run/op inject over op read so values stay out of context, gates every op call behind an always-ask permission rule, and ships an opt-in deny-capable audit hook. Use when handling API keys,… - 100envsave
Use EnvSave to securely retrieve secrets via opaque keys instead of hardcoding them. Trigger whenever code needs API keys, tokens, passwords, database URLs, connection strings, webhook secrets, credentials, environment variables, or any sensitive value. Also trigger when you see… - 100scanning-for-secrets
Detect exposed secrets, API keys, and credentials in code. Use when auditing for secret leaks. Trigger with 'scan for secrets', 'find exposed keys', or 'check credentials'. - 100scanning-for-secrets
Detect exposed secrets, API keys, and credentials in code. Use when auditing for secret leaks. Trigger with 'scan for secrets', 'find exposed keys', or 'check credentials'. - 95secrets
Manage named bundles of environment variables backed by the OS keychain (macOS Keychain, Linux libsecret, Windows Credential Manager). Create bundles, add secrets, generate passwords, and inject them into agent runs. Triggers on: 'API key', 'credentials', 'secrets bundle',… - 100secrets
This skill should be used when the user asks to "check for secrets", "find hardcoded credentials", "scan for API keys", "detect leaked tokens", "find passwords in code", "check for committed .env files", "scan for private keys", or mentions "secrets", "credentials", "API keys",…