BC claudebox
Claude Code running on the network inside a Docker container (aicodebox-based), managed via the claudebox wrapper/CLI. Exposes seven ways to drive it — interactive shell (`claudebox`), one-shot exec (`claudebox "prompt"`), an HTTP REST API (`/run`, async run-id polling, `/files` CRUD), an OpenAI-compatible `/openai/v1/chat/completions` adapter (streaming SSE, multi-turn, multimodal), an MCP server (streamable HTTP, gated by CLAUDEBOX_MCP_MODE — mounts at `/mcp` on the API port when API mode is also on, else runs standalone on its own port), a Telegram bot, and a YAML cron scheduler. Auth is per-mode bearer tokens (CLAUDEBOX_API_MODE_TOKEN, CLAUDEBOX_MCP_MODE_TOKEN, CLAUDEBOX_TELEGRAM_MODE_TOKEN) or none if unset — unset tokens mean the surface is unauthenticated. Use specifically for installing, configuring, launching, or scripting against a claudebox container/wrapper — not as a generic "run any coding task" tool.
Claude Code running on the network inside a Docker container (aicodebox-based), managed via the claudebox wrapper/CLI.
As a process C 55/100 · Has gaps — weak spots: result and completion, inputs and preconditions, running it twice
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The instructions or scripts send data out: environment variables, keys, file contents, chat history. You may never notice, because the agent performs the upload as "part of the task".
If the upload is not needed for the task, remove it: catalogs flag such skills and delist them. If it is needed, name the destination explicitly, say what leaves the machine, and give the user a switch.
How to improve
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 5
✓ No critical or high findings
Medium and low: 5
-
medium Exfiltration
net-credential-useSKILL.md:368Credential used in a network call (verify the destination is the intended service)RUN_ID=$(curl -s -X POST http://loca…080/run -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
-
medium Exfiltration
net-credential-useSKILL.md:371Credential used in a network call (verify the destination is the intended service)until curl -s "http://loca…080/run/result?runId=…" -H "Authorization: Bearer $TOKEN" | jq -e '.status != "running"' >/dev/null; do
-
medium Exfiltration
net-credential-useSKILL.md:374Credential used in a network call (verify the destination is the intended service)curl -s "http://loca…080/run/result?runId=…" -H "Authorization: Bearer $TOKEN" | jq
-
low Dangerous commands
cmd-privilegereferences/setup.md:80Privilege escalation / world-writable permissions (detector / deny-list definition)- `--security-opt no-new-privileges:true` — block setuid privilege escalation inside the container.
detector -
low Risky intent
intent-offensive-securityreferences/setup.md:80Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)- `--security-opt no-new-privileges:true` — block setuid privilege escalation inside the container.
detector
Files scanned: 3. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
body-longSKILL.md body ≈ 5942 tokens (recommended < 5000); move details to references/ - note
frontmatter-keyunknown frontmatter key "homepage"
Process rating: all ten parameters 55/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 30Running it twice. 29 mutating operations with no state check
- 50Failures and branches. 0 branches, has a failure section
- 60Tools and files. Uses tools (bash, web) that frontmatter does not declare
- 70When it triggers. States when to use, but not when not to
- 70Execution cost. Instruction body is 5942 tokens
- 100Steps. 24 steps
- 100Consistency. Name and required fields are in place
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 12 top-level sections: this looks like several domains in one skill
- low The response is described with custom markup (4 tags): a typed call is more reliable
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +4Description does not say when NOT to use the skill (false activations)
- +3Description length 928: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- -2localhost URLs: will not work for another user
- +1No license
- +2Single-language instructions
- +5Description quotes 2 example trigger phrases
- +4Structure: 18 headings
- +3Step-by-step instructions: 24 items
- +4Has examples (24 code blocks)
- +4Reference files are cited in the instructions (1 of 1)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 77.