{"_id":"@booga/bspector","_rev":"2-a1e91c520159cb9a741d84397b427dee","name":"@booga/bspector","dist-tags":{"latest":"0.2.0"},"versions":{"0.1.1":{"name":"@booga/bspector","version":"0.1.1","keywords":["cli","agent","prompt","lookup"],"license":"MIT","_id":"@booga/bspector@0.1.1","maintainers":[{"name":"dreaktor","email":"me@dreaktor.com"}],"homepage":"https://github.com/bvasilenko/bSpector#readme","bugs":{"url":"https://github.com/bvasilenko/bSpector/issues"},"bin":{"bspector":"bin/bspector.js"},"dist":{"shasum":"828593b6b80fd4211732ad95b07d6b74bc57d1fb","tarball":"https://registry.npmjs.org/@booga/bspector/-/bspector-0.1.1.tgz","fileCount":5,"integrity":"sha512-HyXWymDRmWx/2HO43bBpho0WdCJnIf9q2KIT1fNXKgB3CETjhAighzJR0M3HUp5w1FDsaI/JLYhQah4YE24zyA==","signatures":[{"sig":"MEQCIG/9+NiG3ZiW13w7PBnlGCZB+x1Qt3Y7qs/mpsuiweofAiALXrsVGmFEwDb4r7Pekk9XYKhFTBQdRH9qjv7in1eesg==","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":4902},"gitHead":"d8fd1fb29ceb4381f5afd1232ea71f9168303ab7","scripts":{"postinstall":"node ./bin/install.js"},"_npmUser":{"name":"dreaktor","email":"me@dreaktor.com"},"repository":{"url":"git+https://github.com/bvasilenko/bSpector.git","type":"git"},"_npmVersion":"10.9.3","description":"Prompt lookup tool. Agent names a vulnerability pattern from a fixed list; bspector returns the prompt for that vulnerability pattern. The prompt tells the agent how to check the skill manifest for that vulnerability.","directories":{},"_nodeVersion":"22.19.0","_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/bspector_0.1.1_1781269654696_0.32313997019259477","host":"s3://npm-registry-packages-npm-production"}},"0.2.0":{"name":"@booga/bspector","version":"0.2.0","description":"Prompt lookup tool. Agent names a vulnerability pattern from a fixed list; bspector returns the prompt for that vulnerability pattern. The prompt tells the agent how to check the skill manifest for that vulnerability.","bin":{"bspector":"bin/bspector.js"},"scripts":{"postinstall":"node ./bin/install.js"},"keywords":["cli","agent","prompt","lookup"],"license":"MIT","repository":{"type":"git","url":"git+https://github.com/bvasilenko/bSpector.git"},"_id":"@booga/bspector@0.2.0","gitHead":"71684d69d35714a1b59d5a479538c7bf942e99ff","bugs":{"url":"https://github.com/bvasilenko/bSpector/issues"},"homepage":"https://github.com/bvasilenko/bSpector#readme","_nodeVersion":"22.19.0","_npmVersion":"10.9.3","dist":{"integrity":"sha512-yGuUnCMw3jsMMeScSBf8L5lmjtYPg+CMzmioDgMbpj9l6xJ788dabg67XROMYDV0QsoCZ8KUcO/xKe6wBKpw3w==","shasum":"13fcf56a8595a1403a57ce910134fade6f123f8a","tarball":"https://registry.npmjs.org/@booga/bspector/-/bspector-0.2.0.tgz","fileCount":5,"unpackedSize":5993,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEQCIH+hPd0ZvWUGsRlYhdXdjXyWtj6TkNzEtmd6lTha3lJ5AiAZATeqxz/tlvcBiY1enohMVxSAugtZAkZSZQVJxUgp3A=="}]},"_npmUser":{"name":"dreaktor","email":"me@dreaktor.com"},"directories":{},"maintainers":[{"name":"dreaktor","email":"me@dreaktor.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/bspector_0.2.0_1783023599506_0.3739785700781082"},"_hasShrinkwrap":false}},"time":{"created":"2026-06-12T13:07:34.530Z","modified":"2026-07-02T20:19:59.754Z","0.1.1":"2026-06-12T13:07:34.861Z","0.2.0":"2026-07-02T20:19:59.650Z"},"bugs":{"url":"https://github.com/bvasilenko/bSpector/issues"},"license":"MIT","homepage":"https://github.com/bvasilenko/bSpector#readme","keywords":["cli","agent","prompt","lookup"],"repository":{"type":"git","url":"git+https://github.com/bvasilenko/bSpector.git"},"description":"Prompt lookup tool. Agent names a vulnerability pattern from a fixed list; bspector returns the prompt for that vulnerability pattern. The prompt tells the agent how to check the skill manifest for that vulnerability.","maintainers":[{"name":"dreaktor","email":"me@dreaktor.com"}],"readme":"# bspector\n\nFor agents that install tools mid-session. Before your AI coding agent runs `npm install some-cli` or registers an MCP server, bspector reads the tool's capability manifest (what files it can touch, what network it calls, what permissions it requests) and decides SAFE, UNSAFE, or MALFORMED. Excessive-agency declarations (e.g. write-everywhere, full network egress) get refused with a directive that tells the agent why and what to try instead. The capability taxonomy is closed (12 vulnerability patterns at v0.1); the prompt library that powers each verdict can evolve through empirical evaluation. Version 0.2.0 embeds a signed fixture corpus that is hand-authored seed material until an evolved corpus ships in a later release.\n\n\nPrompt lookup tool. Agent names a vulnerability pattern from a fixed list; bspector returns the prompt for that vulnerability pattern. The prompt tells the agent how to check the skill manifest for that vulnerability.\n\nBuilt for agentic loops. Scans a skill manifest or extension descriptor, matches against a closed 12-pattern vulnerability taxonomy, writes a verdict directive on stdout, exits with a discriminating code so the calling agent can branch.\n\n```\nbspector scan <artefact>    scan a manifest against the vulnerability-pattern taxonomy; exit 0 / 1 / 2 / 64\nbspector vuln-patterns      list the 12 supported vulnerability-pattern identifiers\nbspector update             self-update to the latest published version\n```\n\nExit code contract: `0` safe (no vulnerability matched), `1` unsafe (vulnerability matched), `2` internal error, `64` malformed input.\n\n## Install\n\n```sh\ncargo install --git https://github.com/bvasilenko/bSpector\n```\n\n## Use\n\n```sh\nbspector scan ./skill.md --artefact-shape skill-md\n# stdout: Inspect the skill markdown as a pre-install capability descriptor before allowing it into the agent environment. ...\n# exit: 1\n\nbspector scan https://example.com/skill.md --artefact-shape skill-md --strictness strict\n# exit: 1\n\nbspector vuln-patterns\n# stdout: prompt-injection\n#         data-exfiltration\n#         ...\n```\n\nOptional flags for `scan`: `--artefact-shape <shape>`, `--strictness <level>`, `--llm-stage`, `--osv-online`, `--manifest <path>`, `--json`, `--quiet`, `--reason <text>`.\n\n`BSUITE_UPDATE_BASE_URL` overrides the signed-manifest update base URL used by `bspector update`.\n\n## Fixture corpus\n\nVersion 0.2.0 embeds `bspector-v0`, a signed hand-authored fixture corpus. It is seed material for command-line integration and contract testing, not production trust material.\n\n## Vulnerability pattern taxonomy\n\nClosed 12-variant `VulnerabilityPattern` enum. The taxonomy is fixed at this version; widening lands in a later version.\n\n| Category | Variants |\n|---|---|\n| Prompt integrity | `prompt-injection`, `system-prompt-leakage`, `memory-poisoning`, `trigger-abuse` |\n| Data and access | `data-exfiltration`, `privilege-escalation`, `least-privilege-violation` |\n| Supply and tools | `supply-chain`, `tool-misuse`, `excessive-agency` |\n| Agent behavior | `rogue-agent`, `output-handling` |\n\n`bspector vuln-patterns` prints the full list.\n\n## License\n\nMIT.\n","readmeFilename":"README.md"}