Belay is an open-source, local-first security layer for AI coding agents (Claude Code, Codex, Cursor, OpenClaw, Hermes Agent and MCP) that blocks dangerous commands, secret leaks, and prompt injection at the tool-call boundary in under 100ms — no LLM in the decision path by default, no cloud, no phone-home.
- capability exposure inferred + 10
- recent drift inferred + 12
- tool safety inferred + 12
- trust mitigators mixed − 3
inferred mixed
The A–E grade is our heuristic synthesis — a "review this" prompt, not a verdict. Each factor is tagged by what backs it: attested (a verifiable record), reported (a third party's claim), or inferred (our own heuristic, e.g. permissions). See methodology.
graded 9m ago · see ecosystem CVEs →
No known CVEs for this server.
- high dangerous code
committed secret: GitHub token
analyzed commit abaff79 · analyzer v28 · 2h ago
skills & prompt files 21
- skill SECBLOK-belay-abaff79/scanner/tests/corpus_scan/benign/hello_skill/SKILL.md
- skill SECBLOK-belay-abaff79/scanner/tests/corpus_scan/malicious/pipe_skill/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign/hello_skill/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_combined/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_design_skill/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_high_accumulation/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_security_doc_prose/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_ssh_pubkey/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_stdlib_import/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_undeclared_caps/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/benign_unpinned_pip/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/edge_credential_read_only/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/edge_malicious_prose_directive/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/edge_real_typosquat/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious/snooping_claude_config_skill/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious/underdeclared_net_skill/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious_cloud_metadata/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious_conversation_exfil/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious_credential_exfil/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious_remote_exec/SKILL.md
- skill SECBLOK-belay-abaff79/skillscan/tests/corpus/malicious_remote_exec_wrapped/SKILL.md
danger signals1
- committed secret GitHub token SECBLOK-belay-abaff79/daemon/src/observe/secrets.rs :152
ghp_01…(40 chars, redacted)
- recent drift +12 capability drift →
[](https://mcpobservatory.com/servers/github:SECBLOK/belay/security) Heuristic, inferred signals — false positives (legitimately powerful tools, forks, language ports) are expected. Treat each as "review this", not a verdict. See the ecosystem-wide picture on the security hub, or the fleet security of SECBLOK.