Users on exec-based clouds (Fly, Render, Koyeb, Northflank, Railway,
Modal, Daytona, E2B, CodeSandbox, GitHub Codespaces) got no warning
when their session ended that their service was still running and
incurring charges. This adds:
- _show_exec_post_session_summary() in shared/common.sh for non-SSH
providers that use CLI exec commands instead of direct SSH
- SPAWN_DASHBOARD_URL for all 10 exec-based clouds so users get
actionable dashboard links
- Post-session summary calls in each cloud's interactive_session()
- 33 new tests covering the exec post-session summary feature
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Unquoted `<< EOF` heredocs in nanoclaw .env file creation cause shell
expansion of the API key value. If an API key contains `$`, backticks,
or `\`, the value is silently corrupted or could trigger command
execution. Replace with `printf '%s'` which safely writes the value
without interpretation.
Also fix unquoted variable expansion in upload_config_file's mv command
and the github-codespaces/openclaw.sh config heredoc.
Fixes 34 scripts across all cloud providers.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace fragile blocklist validation and printf '%q' escaping in upload_file()
with strict allowlist regex [a-zA-Z0-9/_.~-]+ across all non-SSH cloud providers.
For codesandbox, additionally migrate from shell command interpolation to SDK
filesystem API via environment variables, eliminating the injection surface entirely.
Affected clouds: codesandbox, daytona, e2b, fly, koyeb, modal, northflank,
railway, render, sprite
Fixes#989
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add log_install_failed helper to shared/common.sh that provides
structured troubleshooting for agent install failures: possible causes,
SSH debug command (when server IP available), manual install command,
and re-run suggestion. Also improve SSH key registration error message.
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replaces unsafe direct shell interpolation of OPENROUTER_API_KEY with
the inject_env_vars_ssh/inject_env_vars_local helpers that use
single-quoted values to prevent shell injection.
Affected scripts:
- codesandbox/codex.sh
- codesandbox/interpreter.sh
- codesandbox/gptme.sh
- atlanticnet/goose.sh
This is the same class of fix applied in PR #937 to 3 other scripts,
but these 4 were missed.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three scripts were missed in the previous inject_env_vars migration
(commit d785571): atlanticnet/amazonq.sh, codesandbox/cline.sh, and
civo/continue.sh. These still used raw heredoc/echo patterns that
embed OPENROUTER_API_KEY directly into shell command strings without
escaping, allowing potential command injection if the API key contained
shell metacharacters (quotes, backticks, dollar signs).
Replace with inject_env_vars_ssh (Atlantic.Net, Civo) and
inject_env_vars_local (CodeSandbox) which use generate_env_config
to safely single-quote all values.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: Add codesandbox/gptme
Agent: gap-filler
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* fix: use safe quoting pattern for API key injection in gptme
Agent: pr-maintainer
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: B (Discovery Team) <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
- Fix macOS compatibility bug in Atlantic.Net API signature: `base64 -w 0`
fails on macOS (no `-w` flag); add fallback like other providers
- Replace misleading "Use 'csb' CLI dashboard" in CodeSandbox interactive
session with link to the actual web terminal at codesandbox.io/dashboard
- Soften preflight credential check prompt from "will likely fail" to
"will prompt you to authenticate" (scripts have built-in auth flows)
- Bump CLI version to 0.2.72
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Replace unsafe heredoc/echo patterns with inject_env_vars_ssh (Atlantic.Net)
and inject_env_vars_local (CodeSandbox) for API key injection. The previous
patterns embedded OPENROUTER_API_KEY values directly into shell command strings
without escaping, allowing potential command injection if the API key contained
shell metacharacters (quotes, backticks, dollar signs).
Affected scripts (11 total):
- atlanticnet: codex, continue, gemini, gptme, kilocode, opencode
- codesandbox: amazonq, gemini, goose, opencode, plandex
The safe helpers (generate_env_config) properly single-quote all values and
escape embedded single quotes, preventing shell interpretation of special chars.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The file had 5 nearly-identical inline Node.js scripts, each repeating
the same require/init/async-IIFE/try-catch boilerplate. Extract two
shared helpers:
- _csb_sdk_eval: runs arbitrary JS with an authenticated SDK instance
- _csb_run_cmd: connects to a sandbox and runs a command (used by
run_server and interactive_session)
interactive_session was a verbatim copy of run_server — it now delegates
to run_server directly.
Net result: -96 lines, +49 lines (287 → 241 lines total).
Agent: complexity-hunter
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
Replace technical "Remediation steps:" with "How to fix:" and
"Remediation: Check <url>" with "Check your dashboard: <url>" across
14 cloud providers for clearer error guidance. Add actionable error
messages to Atlantic.Net create_server and SSH key registration failures.
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Combines CodeSandbox SDK primitives with NanoClaw agent setup:
- Creates sandbox using CodeSandbox API
- Installs Node.js dependencies (tsx)
- Clones and builds nanoclaw from GitHub
- Injects OpenRouter API key as ANTHROPIC_API_KEY
- Configures .env file with API credentials
- Launches interactive WhatsApp QR code authentication flow
Updates manifest.json matrix status to "implemented"
Agent: gap-filler
Co-authored-by: B (Discovery Team) <6723574+louisgv@users.noreply.github.com>
Add CodeSandbox as a new sandbox cloud provider for running AI agents.
CodeSandbox features:
- Firecracker microVMs with ~2 second start times
- SDK/CLI-based exec (no SSH)
- Free tier: 40 hours/month on Build plan
- Secure isolated environments
Implementation:
- Created codesandbox/lib/common.sh with SDK wrapper functions
- Implemented 3 agent scripts: claude, aider, openclaw
- Added CodeSandbox to manifest.json clouds
- Created matrix entries (3 implemented, 12 missing)
- Updated test/record.sh to list as non-recordable CLI cloud
- Added codesandbox/README.md with usage instructions
The implementation follows the existing pattern from e2b and modal,
using Node.js SDK (@codesandbox/sdk) for sandbox lifecycle management.
Agent: cloud-scout
Co-authored-by: B (Discovery Team) <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
